Laser additive manufacturing dynamic monitoring method and system based on multi-sensor cooperative sensing

Abstract

The invention provides a laser additive manufacturing dynamic monitoring method and system based on multi-sensor cooperative sensing, and the method comprises the steps: S1, constructing a distributed clock synchronization system, and collecting a multi-physical field signal in a laser powder feeding additive manufacturing process; s2, key process parameters in the signals are extracted, time domain alignment is carried out on the key process parameters and pose data of an integrated machine tool motion control system, and time-space associated data are obtained; s3, a space-time association database is established to store space-time association data, and an additive manufacturing three-dimensional model is obtained; s4, the key process parameters are mapped to the additive manufacturing three-dimensional model for visual rendering, and digital twin bodies are generated; and S5, dynamic monitoring and feedback are conducted on the additive manufacturing process based on the digital twin body. Through collaborative innovation of a multi-source heterogeneous data space-time alignment mechanism, a process parameter dynamic compensation algorithm and a physical field visualization engine, precise mapping and real-time twinborn feedback of full-dimensional parameters in the additive manufacturing process are achieved.

Description

Technical Field

[0001] This invention relates to the field of additive manufacturing and intelligent measurement and control interaction technology, specifically to a dynamic monitoring method and system for laser additive manufacturing based on multi-sensor collaborative perception. Background Technology

[0002] The complex light-powder-gas interaction mechanism, poor solid-liquid transformation stability of the molten pool, and numerous influencing factors in laser powder-feed additive manufacturing result in poor repeatability of the final formed parts and difficulty in tracing the location and cause of defects, hindering the large-scale mass production application of laser powder-feed additive manufacturing technology. Existing technologies can monitor the laser powder-feed additive manufacturing process using a single sensor, but this monitoring method often lacks a comprehensive reflection of the process. Furthermore, the absence of a digital twin system for the manufacturing process provides insufficient assistance in tracing the source of manufacturing defects. Therefore, simultaneously acquiring diverse and heterogeneous data, such as coaxial / paraaxial image information from the laser cladding head, molten pool temperature information, and acoustic signals during the process, can more comprehensively reflect the manufacturing process. Constructing a distributed clock synchronization system to establish a time-series-based heterogeneous data fusion architecture and forming a visualized digital twin model of the additive component process information will ultimately lead to a dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative perception. This method can reflect the stability, consistency, and anomalies of the manufacturing process while completing the additive manufacturing of parts.

[0003] Patent document CN113588074B discloses an online monitoring device and defect diagnosis method for LDED based on multi-element optical information of molten pool. Three fiber optic probes move synchronously with the LDED laser head through a synchronization mechanism. The current and voltage output by the photodetector are used to diagnose defects, and the type of defect is determined based on the abnormal signal. Patent document CN117491503A discloses a method for monitoring and detecting defects in laser powder feeding additive manufacturing process based on acoustic emission sensing. The above documents use one or more sensors of the same type, and although the amount of data information is large, the use of a single photoelectric sensor is somewhat lacking in reflecting the comprehensiveness of the manufacturing process.

[0004] Patent document CN120084885A discloses a laser additive manufacturing pore defect monitoring method based on a multi-sensor fusion network. It acquires laser powder bed melting process signals collected by multiple air-borne acoustic emission sensors with different resonant frequencies, and uses a multi-source dataset to train a multi-source sensor adaptive excitation fusion convolutional neural network algorithm to obtain pore defect monitoring results. However, this patent document improves the recognition accuracy by increasing the number of single-type sensors in additive manufacturing, and is geared towards the more stable and environmentally friendly laser powder bed melting process. For high-power laser powder feeding additive manufacturing, single-sensor methods are difficult to decouple the signal noise caused by machine tool movement and powder collision.

[0005] Patent document CN119714143B discloses a method for detecting porosity defects in arc-fuse additive manufacturing based on multi-sensor signals. This method, targeting the arc-fuse additive manufacturing process, acquires high-speed CCD image information, current and voltage signals, and spectral signals in real time. It obtains the size and location information of porosity defects by performing CT scanning on the printed part and correlates this information with the acquired high-speed CCD images, current and voltage signals, and spectral signals. Finally, it optimizes the arc-fuse additive manufacturing process based on the porosity defect feature dataset. However, this patent document focuses on non-destructive and rapid detection of porosity defects in arc-fuse additive manufacturing. In laser-fuse additive manufacturing, the melt pool size is more precise, and powder obstruction is more severe, requiring the integration of more sensitive sensing methods that overcome powder obstruction. Furthermore, a visual digital twin model of the additive manufacturing process is not established. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the purpose of this invention is to provide a dynamic monitoring method and system for laser additive manufacturing based on multi-sensor collaborative sensing.

[0007] A dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative sensing, provided by the present invention, includes: Step S1: Construct a distributed clock synchronization system to collect multi-physics field signals during the laser powder feeding additive manufacturing process; the multi-physics field signals include optical radiation field, temperature field and acoustic emission field signals; Step S2: Extract key process parameters from the multiphysics field signal and align them in the time domain with the pose data of the integrated machine tool motion control system to obtain spatiotemporal correlation data; the key process parameters include molten pool geometry, temperature field distribution, real-time working distance, and acoustic emission signal characteristics; Step S3: Establish a spatiotemporal correlation database to store spatiotemporal correlation data, and further obtain the additive manufacturing 3D model; Step S4: Map the key process parameters to the additive manufacturing 3D model and perform visualization rendering to generate a digital twin; Step S5: Dynamically monitor and provide feedback on the additive manufacturing process based on the digital twin.

[0008] Preferably, step S1 includes the following sub-steps: Step S1.1: Establish a distributed clock synchronization system based on the IEEE 1588 precision time protocol to construct a unified timing benchmark for the entire laser powder feeding additive manufacturing process; Step S1.2: Real-time acquisition of multi-physics field signals during laser powder feeding additive manufacturing process using a multi-physics field sensing array; the multi-physics field sensing array includes a coaxial high-speed camera, a dual-band infrared pyrometer, and a broadband acoustic emission sensor, and the acquisition of multi-source signals is synchronized by hardware triggering.

[0009] Preferably, step S2 includes the following sub-steps: Step S2.1: Construct a multimodal signal feature engineering processing chain; Step S2.2: Integrate the EtherCAT bus data of the machine tool motion control system to capture the spatial coordinates (X,Y,Z) and Euler angle attitude (α,β,γ) of the cladding head in the machine tool coordinate system in real time, and realize the time domain alignment of coordinate data and process parameters through deterministic network transmission protocol.

[0010] Preferably, step S2.1 includes the following sub-steps: Step S2.1.1: Perform Gaussian-Laplace multi-scale decomposition on the video stream of the coaxial high-speed camera, extract the geometric features of the molten pool through fixed threshold segmentation, and establish a dynamic evolution model of the width, length and area of ​​the molten pool; Step S2.1.2: Based on the principle of dual-color thermometry, the temperature field distribution of the molten pool is calculated by the ratio of the two-band radiation intensity, and the temperature gradient is obtained by fitting using the nonlinear least squares method. Step S2.1.3: Perform SURF feature point detection and matching on the off-axis high-speed vision data, calculate the three-dimensional pose relationship between the cladding head nozzle and the forming surface through the perspective n-point algorithm, and output the real-time working distance; Step S2.1.4: Perform bandpass filtering, data windowing, and time-frequency feature extraction on the acoustic emission signal to extract typical time-domain and frequency-domain features of the characteristic frequency band.

[0011] Preferably, step S2.2 includes: Configure a laser power state machine to trigger a coordinate acquisition interrupt service routine when a laser switch-off signal is detected, and perform a continuity check of the coordinate sequence after the laser is retried.

[0012] Preferably, step S3 includes the following sub-steps: Step S3.1: Construct a heterogeneous data fusion architecture based on time series, use the lightweight relational database SQLite3 to realize real-time storage and retrieval of multi-source time series data, and establish a spatiotemporal relational database containing timestamps, process parameter sets and motion coordinate sets; Step S3.2: The dynamic point cloud reconstruction algorithm based on the moving least squares method converts discrete spatiotemporal coordinates into continuous surface topology, and generates a sub-voxel precision additive manufacturing 3D model through Poisson surface reconstruction.

[0013] Preferably, step S3.2 includes: A model for reconstructing the interlayer topography was established. The pose parameters of the cladding head (X,Y,Z,α,β,γ) were used as the reference coordinate system. Based on the real-time monitored working distance deviation Δh, the layer deposition height was dynamically compensated through inverse kinematics calculation.

[0014] Preferably, step S4 includes: The nonlinear mapping from process parameters to the HSV color space is implemented based on the OpenGL shading language. The Phong lighting model is used for parametric field rendering to construct a digital twin with multi-resolution LOD levels, and to visualize and provide feedback on additive manufacturing process variables.

[0015] Preferably, step S4 further includes constructing a parametric field interpolation engine based on CUDA parallel computing, and using radial basis function kernels to realize spatial interpolation of multiphysics fields.

[0016] A dynamic monitoring system for laser additive manufacturing based on multi-sensor collaborative sensing, provided by the present invention, includes: Module M1: Constructs a distributed clock synchronization system to collect multi-physics field signals during the laser powder feeding additive manufacturing process; the multi-physics field signals include optical radiation field, temperature field, and acoustic emission field signals; Module M2: Extracts key process parameters from multiphysics field signals and aligns them in the time domain with the pose data of the integrated machine tool motion control system to obtain spatiotemporal correlation data; the key process parameters include molten pool geometry, temperature field distribution, real-time working distance, and acoustic emission signal characteristics; Module M3: Establishes a spatiotemporal correlation database to store spatiotemporal correlation data, and further obtains additive manufacturing 3D models; Module M4: Maps the key process parameters to the additive manufacturing 3D model and performs visualization rendering to generate a digital twin; Module M5: Dynamically monitors and provides feedback on the additive manufacturing process based on the digital twin.

[0017] Preferably, module M1 includes the following sub-modules: Module M1.1: Establish a distributed clock synchronization system based on the IEEE 1588 precision time protocol to construct a unified timing reference for the entire laser powder feeding additive manufacturing process; Module M1.2: Real-time acquisition of multi-physics field signals during laser powder feeding additive manufacturing process via a multi-physics field sensing array; the multi-physics field sensing array includes a coaxial high-speed camera, a dual-band infrared pyrometer, and a broadband acoustic emission sensor, and adopts a hardware triggering method to realize the synchronous acquisition of multi-source signals.

[0018] Preferably, module M2 includes the following sub-modules: Module M2.1: Constructing a multimodal signal feature engineering processing chain; Module M2.2: Integrates EtherCAT bus data from the machine tool motion control system, captures the spatial coordinates (X,Y,Z) and Euler angle attitude (α,β,γ) of the cladding head in the machine tool coordinate system in real time, and achieves time-domain alignment of coordinate data with process parameters through a deterministic network transmission protocol.

[0019] Preferably, module M2.1 includes the following sub-modules: Module M2.1.1: Performs Gaussian-Laplace multi-scale decomposition on the video stream of a coaxial high-speed camera, extracts the geometric features of the molten pool through fixed threshold segmentation, and establishes a dynamic evolution model of the width, length and area of ​​the molten pool; Module M2.1.2: Based on the principle of dual-color temperature measurement, the temperature field distribution of the molten pool is calculated by the ratio of the two-band radiation intensity, and the temperature gradient is obtained by fitting using the nonlinear least squares method; Module M2.1.3: Performs SURF feature point detection and matching on the off-axis high-speed vision data, calculates the three-dimensional pose relationship between the cladding head nozzle and the forming surface through the perspective n-point algorithm, and outputs the real-time working distance; Module M2.1.4: Performs bandpass filtering, data windowing, and time-frequency feature extraction on acoustic emission signals, extracting typical time-domain and frequency-domain features of characteristic frequency bands.

[0020] Preferably, the module M2.2 includes: Configure a laser power state machine to trigger a coordinate acquisition interrupt service routine when a laser switch-off signal is detected, and perform a continuity check of the coordinate sequence after the laser is retried.

[0021] Preferably, module M3 includes the following sub-modules: Module M3.1: Construct a heterogeneous data fusion architecture based on time series, use the lightweight relational database SQLite3 to realize real-time storage and retrieval of multi-source time series data, and establish a spatiotemporal relational database containing timestamps, process parameter sets and motion coordinate sets; Module M3.2: A dynamic point cloud reconstruction algorithm based on the moving least squares method, which converts discrete spatiotemporal coordinates into continuous surface topology and generates sub-voxel precision additive manufacturing 3D models through Poisson surface reconstruction.

[0022] Preferably, module M3.2 includes: A model for reconstructing the interlayer topography was established. The pose parameters of the cladding head (X,Y,Z,α,β,γ) were used as the reference coordinate system. Based on the real-time monitored working distance deviation Δh, the layer deposition height was dynamically compensated through inverse kinematics calculation.

[0023] Preferably, the module M4 includes: The nonlinear mapping from process parameters to the HSV color space is implemented based on the OpenGL shading language. The Phong lighting model is used for parametric field rendering to construct a digital twin with multi-resolution LOD levels, and to visualize and provide feedback on additive manufacturing process variables.

[0024] Preferably, module M4 further includes a parameter field interpolation engine based on CUDA parallel computing, which uses radial basis function kernels to realize spatial interpolation of multiphysics fields.

[0025] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention utilizes multiple types of sensors to synchronously monitor the laser powder feeding additive manufacturing process; leveraging the unique monitoring capabilities and characteristics of various sensors, it comprehensively and multidimensionally reflects the dynamic changes in the process; through cross-validation and in-depth analysis of monitoring data from different sources, compared with traditional methods, it can provide more accurate and comprehensive feedback on forming quality, significantly improving the effectiveness and reliability of quality feedback.

[0026] 2. This invention integrates coordinate values ​​and working distance information to obtain surface contour point cloud data that can accurately characterize the actual printed size of the deposited part; using advanced reverse modeling technology, these point cloud data are transformed into a three-dimensional model, which has the ability to reflect the dynamic changes in the size of the printed part in real time, providing a powerful tool for accurately grasping the size evolution during the printing process.

[0027] 3. This invention visually presents key variables in the process in the form of color changes on a three-dimensional model. This visualization strategy makes the feedback effect of quality monitoring more prominent. Operators can intuitively compare and analyze potential defects through the color distribution of the model, providing a convenient and efficient way to adjust process parameters and optimize the manufacturing process in a timely manner.

[0028] 4. This invention achieves accurate mapping and real-time twin feedback of all-dimensional parameters in the additive manufacturing process through the collaborative innovation of a multi-source heterogeneous data spatiotemporal alignment mechanism, a dynamic compensation algorithm for process parameters, and a physical field visualization engine. Attached Figure Description

[0029] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating the method of the present invention.

[0030] Figure 2 This is a schematic diagram of the surface contour point cloud construction method of the present invention. Detailed Implementation

[0031] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0032] A dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative sensing includes: Step S1: Construct a distributed clock synchronization system to collect multi-physics field signals during the laser powder feeding additive manufacturing process; the multi-physics field signals include optical radiation field, temperature field and acoustic emission field signals; Step S2: Extract key process parameters from the multiphysics field signal and align them in the time domain with the pose data of the integrated machine tool motion control system to obtain spatiotemporal correlation data; the key process parameters include molten pool geometry, temperature field distribution, real-time working distance, and acoustic emission signal characteristics; Step S3: Establish a spatiotemporal correlation database to store spatiotemporal correlation data, and further obtain the additive manufacturing 3D model; Step S4: Map the key process parameters to the additive manufacturing 3D model and perform visualization rendering to generate a digital twin; Step S5: Dynamically monitor and provide feedback on the additive manufacturing process based on the digital twin.

[0033] In one embodiment, the above steps specifically include: like Figure 1 As shown, according to the proposed method for dynamic monitoring of laser powder feeding additive manufacturing based on multi-sensor collaborative perception, a distributed clock synchronization system is first constructed. Based on this, multiple sensing devices, including a coaxial optical camera, a rangefinder high-speed camera, a colorimetric pyrometer, and an acoustic emission sensor, are used to collect signals during the laser powder feeding additive manufacturing process, and each collected raw signal is timestamped. Subsequently, algorithms such as Gaussian-Laplace multi-scale decomposition and SURF feature matching are used to extract key process parameters such as molten pool geometry, temperature, and working distance from the raw signals. Simultaneously, for the integrated machine tool motion control system of the additive manufacturing equipment, the spatial coordinates of the cladding head in the machine tool coordinate system are captured in real time. Then, the coordinate data and process parameters are aligned in the time domain, establishing a spatiotemporal correlation database containing timestamps, process parameter sets, and motion coordinate sets. Using the coordinate values ​​and process variables in this database as references, a three-dimensional model of the additively manufactured part is constructed through reverse modeling. Finally, a color scale for process variables is constructed, and the color of each coordinate point in the 3D model is modified according to this scale to ensure that the colors between coordinate points transition in a gradual manner, thereby obtaining a digital twin model that can visualize and provide feedback on the variables of the additive manufacturing forming process.

[0034] like Figure 2 As shown, in the laser powder feeding additive manufacturing dynamic monitoring method based on multi-sensor collaborative sensing provided by this invention, the coordinate values ​​(X, Y, Z) of the cladding head are used as the basic reference. The method integrates the working distance from the cladding head to the printing surface, accurately extracted during multi-sensor monitoring. h performs numerical addition and subtraction operations on the coordinate values ​​(X, Y, Z) along the Z-axis. This operation yields surface contour point cloud data that accurately represents the dynamic changes in the printed dimensions of the actual deposited part. Based on this data, a 3D model of the deposited part is constructed using reverse modeling techniques.

[0035] Specifically, the following steps are included: Step S1: Establish a distributed clock synchronization system based on the IEEE 1588 precision time protocol to build a unified timing reference for the entire laser additive manufacturing process; The time synchronization accuracy of the unified timing reference is δt≤10μs, and the clock drift rate is ρ≤1ppm, which meets the requirements of Nyquist sampling theorem for multi-source signal acquisition. Step S2: Real-time acquisition of light radiation field, temperature field, and acoustic emission field signals during the additive manufacturing process is achieved using a multi-physics field sensor array. The sensor array includes a coaxial high-speed camera, a dual-band infrared pyrometer, and a broadband acoustic emission sensor. Hardware triggering is used to achieve sub-millisecond synchronization of multi-source signal acquisition. In one embodiment, a distributed system is constructed, using the Network Time Protocol (NTP) to synchronize the clocks of multiple data acquisition devices. Specifically, time is synchronized via the protocol on the network interfaces of each card to ensure that data acquisition occurs on the same time base.

[0036] The sensor adopts a composite protection design of opto-isolation and electromagnetic shielding, and its dynamic range meets the following requirements: light radiation intensity 10^3-10^6 W / m², temperature detection range 300-3300K, and sound pressure level 40-120dB. Step S3: Construct a multimodal signal feature engineering processing chain, specifically including: (3a) Gaussian-Laplace multiscale decomposition is performed on the video stream of the coaxial high-speed camera. Geometric features of the molten pool are extracted by segmentation with a fixed threshold, and a dynamic evolution model of the width, length and area of ​​the molten pool is established. (3b) Based on the principle of dual-color thermometry, the temperature field distribution of the molten pool is calculated by the ratio of the two-band radiation intensity, and the temperature gradient is obtained by fitting using the nonlinear least squares method. (3c) Perform SURF feature point detection and matching on the off-axis high-speed vision data, and calculate the three-dimensional pose relationship between the cladding head nozzle and the forming surface through the perspective n-point (PnP) algorithm to output the real-time working distance; (3d) Bandpass filtering, data windowing, and time-frequency feature extraction are performed on the acoustic emission signal to extract typical time-domain and frequency-domain features of the 50~400kHz characteristic frequency band; Step S4: Integrate the EtherCAT bus data of the machine tool motion control system to capture the spatial coordinates (X,Y,Z) and Euler angle attitude (α,β,γ) of the cladding head in the machine tool coordinate system in real time. A deterministic network transmission protocol is used to achieve time-domain alignment between the coordinate data and process parameters. For example, based on the IEEE 1588 unified timestamp reference of a distributed clock synchronization system, a four-step time-domain alignment process of "timestamp embedding - data synchronization transmission - interpolation registration - deviation correction" is executed through a deterministic network transmission protocol to achieve high-precision time-domain alignment between the coordinate data and process parameters, with an overall time registration deviation ≤10μs.

[0037] Configure a laser power state machine. When a laser switch-off signal (TTL low level) is detected, trigger the coordinate acquisition interrupt service routine and perform a continuity check of the coordinate sequence after laser re-triggering. Step S5: Construct a heterogeneous data fusion architecture based on time series, using the lightweight relational database SQLite3 to achieve real-time storage and retrieval of multi-source time series data (sampling rate 1-50kHz), and establish a spatiotemporal correlation database containing timestamp τ, process parameter sets, and motion coordinate sets. In a specific embodiment, it is a spatiotemporal correlation database of "timestamp anchoring - spatial coordinate indexing - multi-dimensional parameter correlation"; this spatiotemporal correlation uses a unified high-precision timestamp τ as the core time axis and the spatial coordinates (X,Y,Z,α,β,γ) of the cladding head as the spatial index dimension, binding process parameters, motion coordinate data, and equipment status parameters at the same time node.

[0038] Step S6: The dynamic point cloud reconstruction algorithm based on the moving least squares method converts discrete spatiotemporal coordinates into continuous surface topology, and generates a sub-voxel precision additive manufacturing 3D model through Poisson surface reconstruction. A model for reconstructing the interlayer topography was established. The pose parameters of the cladding head (X,Y,Z,α,β,γ) were used as the reference coordinate system. Based on the real-time monitored working distance deviation Δh, the layer deposition height was dynamically compensated through inverse kinematics calculation.

[0039] The specific process is as follows: 1. Construct the coordinate system of the cladding head: with the center of the cladding head nozzle as the origin, the Z-axis points along the axial direction of the cladding head towards the workpiece, the X-axis is along the direction of travel of the cladding head, and the Y-axis is determined by the right-hand rule; 2. Calculate the theoretical deposition height: Basic process experiments need to be carried out in the early stage to calibrate the relationship between laser process parameters and deposition height; 3. Introduce actual working distance deviation: The working distance of the cladding head is monitored in real time, and the actual working distance deviation is calculated. For planar deposition, only the deposition height deviation in the Z-axis direction needs to be supplemented; for complex curved surfaces, the actual working distance deviation needs to be decomposed into the normal direction of the tool coordinate system, thereby realizing the calculation of the deposition height compensation value.

[0040] Step S7: Implement nonlinear mapping from process parameters to HSV color space based on OpenGL shading language, use Phong lighting model for parametric field rendering, construct a digital twin with multi-resolution LOD levels, and visualize the additive manufacturing process variables.

[0041] The process of nonlinear mapping includes: A nonlinear normalization function is used to map process parameters (including molten pool geometry, temperature field distribution, real-time working distance, and acoustic emission signal characteristics) to the [0,1] interval, and the normalized values ​​are mapped to the H component (hue, 0-360°) of the HSV color space. This mapping is then performed on each vertex or fragment of the 3D model in the OpenGL fragment shader to achieve real-time rendering.

[0042] A parametric field interpolation engine based on CUDA parallel computing was constructed, and a radial basis function (RBF) kernel was used to realize the spatial interpolation of multiphysics fields. The color gradient transition satisfies the uniformity criterion of the CIE Lab color space.

[0043] CIE Lab is a uniform color space defined by the International Commission on Illumination. In the CIE Lab color space, the straight-line distance between two points—a standard color and a measured sample color—is directly calculated. "Uniformity" means that in the Lab space, the Euclidean distance between two color points is proportional to the color difference perceived by the human eye; therefore, regardless of the position of these colors within the color gamut, the degree of color difference perceived by the human eye is the same.

[0044] The specific implementation process is as follows: First, the process parameters are mapped to the HSV color space to obtain the initial color; then, using RGB as an intermediate space, the HSV color is converted to Lab; uniform interpolation is performed in the Lab space; and finally, the Lab color is converted back to RGB for display.

[0045] Through the above steps, the visual difference between adjacent colors in the final rendered digital twin during the color gradient process is proportional to the numerical difference of the process parameters, which significantly improves the accuracy and intuitiveness of the monitoring personnel's perception of changes in process parameters.

[0046] The present invention also provides a dynamic monitoring system for laser additive manufacturing based on multi-sensor collaborative perception. The dynamic monitoring system for laser additive manufacturing based on multi-sensor collaborative perception can be implemented by executing the process steps of the dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative perception. That is, those skilled in the art can understand the dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative perception as a preferred embodiment of the dynamic monitoring system for laser additive manufacturing based on multi-sensor collaborative perception.

[0047] Specifically, a dynamic monitoring system for laser additive manufacturing based on multi-sensor collaborative perception includes: Module M1: Constructs a distributed clock synchronization system to collect multi-physics field signals during the laser powder feeding additive manufacturing process; The multiphysics field signals include optical radiation field, temperature field and acoustic emission field signals; Module M2: Extracts key process parameters from multiphysics field signals and aligns them in the time domain with the pose data of the integrated machine tool motion control system to obtain spatiotemporal correlation data; the key process parameters include molten pool geometry, temperature field distribution, real-time working distance, and acoustic emission signal characteristics; Module M3: Establishes a spatiotemporal correlation database to store spatiotemporal correlation data, and further obtains additive manufacturing 3D models; Module M4: Maps the key process parameters to the additive manufacturing 3D model and performs visualization rendering to generate a digital twin; Module M5: Dynamically monitors and provides feedback on the additive manufacturing process based on the digital twin.

[0048] The module M1 includes the following sub-modules: Module M1.1: Establish a distributed clock synchronization system based on the IEEE 1588 precision time protocol to construct a unified timing reference for the entire laser powder feeding additive manufacturing process; Module M1.2: Real-time acquisition of multi-physics field signals during laser powder feeding additive manufacturing process via a multi-physics field sensing array; the multi-physics field sensing array includes a coaxial high-speed camera, a dual-band infrared pyrometer, and a broadband acoustic emission sensor, and adopts a hardware triggering method to realize the synchronous acquisition of multi-source signals.

[0049] Module M2 includes the following sub-modules: Module M2.1: Constructing a multimodal signal feature engineering processing chain; Module M2.2: Integrates EtherCAT bus data from the machine tool motion control system, captures the spatial coordinates (X,Y,Z) and Euler angle attitude (α,β,γ) of the cladding head in the machine tool coordinate system in real time, and achieves time-domain alignment of coordinate data with process parameters through a deterministic network transmission protocol.

[0050] Module M2.1 includes the following sub-modules: Module M2.1.1: Performs Gaussian-Laplace multi-scale decomposition on the video stream of a coaxial high-speed camera, extracts the geometric features of the molten pool through fixed threshold segmentation, and establishes a dynamic evolution model of the width, length and area of ​​the molten pool; Module M2.1.2: Based on the principle of dual-color temperature measurement, the temperature field distribution of the molten pool is calculated by the ratio of the two-band radiation intensity, and the temperature gradient is obtained by fitting using the nonlinear least squares method; Module M2.1.3: Performs SURF feature point detection and matching on the off-axis high-speed vision data, calculates the three-dimensional pose relationship between the cladding head nozzle and the forming surface through the perspective n-point algorithm, and outputs the real-time working distance; Module M2.1.4: Performs bandpass filtering, data windowing, and time-frequency feature extraction on acoustic emission signals, extracting typical time-domain and frequency-domain features of characteristic frequency bands.

[0051] The module M2.2 includes: Configure a laser power state machine to trigger a coordinate acquisition interrupt service routine when a laser switch-off signal is detected, and perform a continuity check of the coordinate sequence after the laser is retried.

[0052] The module M3 includes the following sub-modules: Module M3.1: Construct a heterogeneous data fusion architecture based on time series, use the lightweight relational database SQLite3 to realize real-time storage and retrieval of multi-source time series data, and establish a spatiotemporal relational database containing timestamps, process parameter sets and motion coordinate sets; Module M3.2: A dynamic point cloud reconstruction algorithm based on the moving least squares method, which converts discrete spatiotemporal coordinates into continuous surface topology and generates sub-voxel precision additive manufacturing 3D models through Poisson surface reconstruction.

[0053] The module M3.2 includes: A model for reconstructing the interlayer topography was established. The pose parameters of the cladding head (X,Y,Z,α,β,γ) were used as the reference coordinate system. Based on the real-time monitored working distance deviation Δh, the layer deposition height was dynamically compensated through inverse kinematics calculation.

[0054] The module M4 includes: The nonlinear mapping from process parameters to the HSV color space is implemented based on the OpenGL shading language. The Phong lighting model is used for parametric field rendering to construct a digital twin with multi-resolution LOD levels, and to visualize and provide feedback on additive manufacturing process variables.

[0055] The module M4 also includes a parameter field interpolation engine based on CUDA parallel computing, which uses radial basis function kernels to realize spatial interpolation of multiphysics fields.

[0056] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0057] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative sensing, characterized in that, include: Step S1: Construct a distributed clock synchronization system to collect multi-physics field signals during the laser powder feeding additive manufacturing process; The multiphysics field signals include optical radiation field, temperature field and acoustic emission field signals; Step S2: Extract key process parameters from the multiphysics field signal and align them in the time domain with the pose data of the integrated machine tool motion control system to obtain spatiotemporal correlation data; the key process parameters include molten pool geometry, temperature field distribution, real-time working distance, and acoustic emission signal characteristics; Step S3: Establish a spatiotemporal correlation database to store spatiotemporal correlation data, and further obtain the additive manufacturing 3D model; Step S4: Map the key process parameters to the additive manufacturing 3D model and perform visualization rendering to generate a digital twin; Step S5: Dynamically monitor and provide feedback on the additive manufacturing process based on the digital twin.

2. The dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative sensing according to claim 1, characterized in that, Step S1 includes the following sub-steps: Step S1.1: Establish a distributed clock synchronization system based on the IEEE 1588 precision time protocol to construct a unified timing benchmark for the entire laser powder feeding additive manufacturing process; Step S1.2: Real-time acquisition of multi-physics field signals during laser powder feeding additive manufacturing process using a multi-physics field sensing array; the multi-physics field sensing array includes a coaxial high-speed camera, a dual-band infrared pyrometer, and a broadband acoustic emission sensor, and the acquisition of multi-source signals is synchronized by hardware triggering.

3. The dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative sensing according to claim 1, characterized in that, Step S2 includes the following sub-steps: Step S2.1: Construct a multimodal signal feature engineering processing chain; Step S2.2: Integrate the EtherCAT bus data of the machine tool motion control system to capture the spatial coordinates (X,Y,Z) and Euler angle attitude (α,β,γ) of the cladding head in the machine tool coordinate system in real time, and realize the time domain alignment of coordinate data and process parameters through deterministic network transmission protocol.

4. The dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative sensing according to claim 3, characterized in that, Step S2.1 includes the following sub-steps: Step S2.1.1: Perform Gaussian-Laplace multi-scale decomposition on the video stream of the coaxial high-speed camera, extract the geometric features of the molten pool through fixed threshold segmentation, and establish a dynamic evolution model of the width, length and area of ​​the molten pool; Step S2.1.2: Based on the principle of dual-color thermometry, the temperature field distribution of the molten pool is calculated by the ratio of the two-band radiation intensity, and the temperature gradient is obtained by fitting using the nonlinear least squares method. Step S2.1.3: Perform SURF feature point detection and matching on the off-axis high-speed vision data, calculate the three-dimensional pose relationship between the cladding head nozzle and the forming surface through the perspective n-point algorithm, and output the real-time working distance; Step S2.1.4: Perform bandpass filtering, data windowing, and time-frequency feature extraction on the acoustic emission signal to extract typical time-domain and frequency-domain features of the characteristic frequency band.

5. The dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative sensing according to claim 3, characterized in that, Step S2.2 includes: Configure a laser power state machine to trigger a coordinate acquisition interrupt service routine when a laser switch-off signal is detected, and perform a continuity check of the coordinate sequence after the laser is retried.

6. The dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative sensing according to claim 1, characterized in that, Step S3 includes the following sub-steps: Step S3.1: Construct a heterogeneous data fusion architecture based on time series, use the lightweight relational database SQLite3 to realize real-time storage and retrieval of multi-source time series data, and establish a spatiotemporal relational database containing timestamps, process parameter sets and motion coordinate sets; Step S3.2: The dynamic point cloud reconstruction algorithm based on the moving least squares method converts discrete spatiotemporal coordinates into continuous surface topology, and generates a sub-voxel precision additive manufacturing 3D model through Poisson surface reconstruction.

7. The dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative sensing according to claim 6, characterized in that, Step S3.2 includes: A model for reconstructing the interlayer topography was established. The pose parameters of the cladding head (X,Y,Z,α,β,γ) were used as the reference coordinate system. Based on the real-time monitored working distance deviation Δh, the layer deposition height was dynamically compensated through inverse kinematics calculation.

8. The dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative sensing according to claim 1, characterized in that, Step S4 includes: The nonlinear mapping from process parameters to the HSV color space is implemented based on the OpenGL shading language. The Phong lighting model is used for parametric field rendering to construct a digital twin with multi-resolution LOD levels, and to visualize and provide feedback on additive manufacturing process variables.

9. The dynamic monitoring method for laser additive manufacturing based on multi-sensor collaborative sensing according to claim 8, characterized in that, Step S4 further includes constructing a parametric field interpolation engine based on CUDA parallel computing, and using radial basis function kernels to realize spatial interpolation of multiphysics fields.

10. A dynamic monitoring system for laser additive manufacturing based on multi-sensor collaborative perception, characterized in that, include: Module M1: Constructs a distributed clock synchronization system to collect multi-physics field signals during the laser powder feeding additive manufacturing process; The multiphysics field signals include optical radiation field, temperature field and acoustic emission field signals; Module M2: Extracts key process parameters from multiphysics field signals and aligns them in the time domain with the pose data of the integrated machine tool motion control system to obtain spatiotemporal correlation data; the key process parameters include molten pool geometry, temperature field distribution, real-time working distance, and acoustic emission signal characteristics; Module M3: Establishes a spatiotemporal correlation database to store spatiotemporal correlation data, and further obtains additive manufacturing 3D models; Module M4: Maps the key process parameters to the additive manufacturing 3D model and performs visualization rendering to generate a digital twin; Module M5: Dynamically monitors and provides feedback on the additive manufacturing process based on the digital twin.

Images