A method and system for online reconstruction of three-dimensional shape of a forging and deviation detection
By using a combination of narrowband blue laser and filters during the forging process, along with mold feature surfaces and calibration blocks, a global coordinate system is established, enabling high-precision online reconstruction and deviation detection of the three-dimensional morphology of forgings. This solves the problems of detection lag and low accuracy under high-temperature environments, and supports process optimization and scrap reduction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HENAN ANQIAN INTELLIGENT EQUIP CO LTD
- Filing Date
- 2026-04-30
- Publication Date
- 2026-07-21
AI Technical Summary
Existing forging processes suffer from problems such as detection lag, low accuracy, and high misjudgment rate under high-temperature conditions, making it impossible to achieve high-precision three-dimensional reconstruction and deviation detection of all process nodes of forgings.
Using a 405nm-450nm narrowband blue laser as the active light source, combined with a narrowband filter and a short exposure imaging strategy, high-temperature infrared radiation interference is suppressed. By combining the mold fixed feature surface and the high-temperature resistant calibration block, a global unified coordinate system is established, and the three-dimensional reconstruction and deviation detection of the forging are performed through multi-dimensional comprehensive indicators.
It enables real-time, high-precision detection of the forging process, reduces scrap generation, improves the robustness and adaptability of the detection system, and supports real-time adjustment of process parameters.
Smart Images
Figure CN122435152A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of online inspection technology in forging processing, and discloses a method and system for online reconstruction of the three-dimensional morphology of forgings and deviation detection. Background Technology
[0002] Forging is a core process for the plastic forming of metals and is widely used in key fields such as aerospace, wind power, construction machinery, and automobile manufacturing. The dimensional accuracy and surface forming quality of forgings directly determine the subsequent machining allowance, the service performance of parts, and the production qualification rate.
[0003] Currently, forging quality inspection mainly relies on offline inspection, which involves measuring the dimensions and morphology of forgings after final forging and cooling using coordinate measuring machines (CMMs) or manual calipers. This method suffers from significant delays, failing to detect excessive pre-forging deformation or forming defects during the forging process. By the time defective products are detected, substantial waste of raw materials, energy, and time has occurred, especially for large, precision forgings, where the loss per scrap is extremely high. Furthermore, offline inspection cannot track the deformation process of forgings at all stages—before, after, and after final forging—making it difficult to support iterative optimization of the forming process.
[0004] Existing online inspection solutions are limited. Forging sites are subject to strong interference from high-temperature forgings (above 1000℃), oxide scale splashing, and environmental vibration. Ordinary visible light vision and conventional laser inspection solutions are easily affected by high-temperature infrared radiation, resulting in noisy point cloud data and missing effective signals, making it impossible to achieve high-precision 3D reconstruction. Furthermore, the lack of a globally unified coordinate reference during multi-process node and multi-camera acquisition processes easily leads to coordinate offsets, resulting in insufficient accuracy in deviation calculation. Summary of the Invention
[0005] To address the aforementioned shortcomings of existing technologies, the present invention aims to provide a method and system for online reconstruction and deviation detection of three-dimensional morphology of forgings. This method and system can overcome the strong interference in high-temperature forging environments, achieve high-quality point cloud acquisition and high-precision three-dimensional reconstruction of all process nodes of forgings, accurately determine forming deviations through multi-dimensional comprehensive indicators, and have an adaptive threshold adjustment function. This solves the problems of lagging offline detection, low acquisition accuracy in high-temperature environments, single detection dimensions, and high misjudgment rate in existing technologies.
[0006] To achieve the above-mentioned technical effects, the technical solution adopted by the present invention is: a method for online reconstruction and deviation detection of three-dimensional morphology of forgings, comprising the following steps: Point cloud data of the forging was collected at multiple process nodes before pre-forging, after pre-forging, and after final forging. Obtain the fixed feature surfaces of the mold and the feature point cloud of the high-temperature resistant calibration block on site, and register the feature point cloud with the preset reference model to establish a globally unified fixed coordinate system; In the fixed coordinate system, the point cloud data is denoised and rigid body registered to obtain a three-dimensional reconstruction model of the forging. The three-dimensional reconstruction model is spatially aligned with a preset standard model to obtain local shape deviations and volume deviations. Based on the local morphological deviations and volume deviations, a comprehensive evaluation index is calculated using a comprehensive judgment function. The comprehensive evaluation index is compared with the set threshold. When the set threshold is exceeded, it is determined that the forging is deformed excessively and an alarm signal is output.
[0007] In a preferred embodiment, the mold is a fixing device in a forging production line for forming the forging, and the fixing feature surface of the mold is a geometric feature surface on the mold whose position remains unchanged during the forging process; The point cloud data is collected using a high-temperature interference resistance strategy, specifically including: The surface of the forging is irradiated by emitting a narrow-band blue laser with a wavelength range of 405nm-450nm as an active light source. The laser reflection signal is received by an industrial camera equipped with a narrowband filter; By employing short-exposure imaging and delayed acquisition strategies, high-temperature background radiation is suppressed, resulting in high-quality point cloud data.
[0008] As a preferred embodiment, obtaining the local morphological deviation specifically includes: The local topography deviation value is obtained by calculating the Euclidean distance between each point in the current point cloud and the corresponding point in the standard model; The local curvature index is obtained by calculating the change in the normal vector between a point and its neighboring points. Based on the local morphological deviation value and the local curvature index, points that simultaneously meet the deviation threshold condition and the curvature threshold condition are selected to obtain the defect area.
[0009] As a preferred embodiment, the expression of the comprehensive determination function is: ; Where J is the comprehensive judgment index, For the maximum local deviation, Here, V represents the standard deviation of the deviation, and V represents the current volume. Where D is the theoretical volume and D is the characteristic dimension of the forging. Where N is the number of defect points, and N is the total number of points. , , , These are the weighting coefficients.
[0010] As a preferred implementation, the method also includes threshold adaptive updating, specifically including: The sample distribution characteristics are obtained by statistically analyzing the mean and standard deviation of the deviation indicators of historical qualified samples. The expression for threshold adaptive update is: ; Among them, T new The updated threshold To update the step size, T old The old threshold =2, This represents the mean of the deviation index.
[0011] Based on the above method, the present invention also proposes an online reconstruction and deviation detection system for the three-dimensional morphology of forgings, comprising: The data acquisition module is used to collect point cloud data of the forging at multiple process nodes before pre-forging, after pre-forging, and after final forging. The coordinate establishment module is used to acquire the fixed feature surfaces of the mold and the feature point cloud of the high-temperature resistant calibration block on site, and to register the feature point cloud with the preset reference model to establish a globally unified fixed coordinate system. The reconstruction processing module is used to denoise and rigidly register the point cloud data in the fixed coordinate system to obtain a three-dimensional reconstruction model of the forging. The deviation calculation module is used to spatially align the three-dimensional reconstructed model with a preset standard model to obtain local topographic deviations and volume deviations. The evaluation and judgment module is used to calculate the comprehensive evaluation index based on the local morphological deviation and volume deviation through a comprehensive judgment function. The alarm module is used to compare the comprehensive evaluation index with the set threshold. When the set threshold is exceeded, it is determined that the forging deformation is excessive and an alarm signal is output.
[0012] In a preferred embodiment, the data acquisition module is specifically used for: The surface of the forging is irradiated by emitting a narrow-band blue laser with a wavelength range of 405nm-450nm as an active light source. The laser reflection signal is received by an industrial camera equipped with a narrowband filter; High-quality point cloud data was obtained by suppressing high-temperature background radiation through short-exposure imaging and delayed acquisition strategies.
[0013] In a preferred embodiment, the deviation calculation module is specifically used for: The local topography deviation value is obtained by calculating the Euclidean distance between each point in the current point cloud and the corresponding point in the standard model; The local curvature index is obtained by calculating the change in the normal vector between a point and its neighboring points. Based on the local morphological deviation value and the local curvature index, points that simultaneously meet the deviation threshold condition and the curvature threshold condition are selected to obtain the defect area.
[0014] In a preferred embodiment, the expression of the comprehensive judgment function in the evaluation and judgment module is as follows: ; Where J is the comprehensive judgment index, For the maximum local deviation, Here, V represents the standard deviation of the deviation, and V represents the current volume. Where D is the theoretical volume and D is the characteristic dimension of the forging. Where N is the number of defect points, and N is the total number of points. , , , These are the weighting coefficients.
[0015] As a preferred implementation, it also includes a threshold update module, which is used to obtain the sample distribution characteristics by statistically analyzing the mean and standard deviation of the deviation indicators of historical qualified samples, and to perform adaptive threshold updates. The expression for the adaptive threshold update is: ; Among them, T new The updated threshold To update the step size, T old The old threshold =2, This represents the mean of the deviation index.
[0016] Compared with existing technologies, this invention uses a 405nm-450nm narrowband blue laser as the active light source, combined with a matching narrowband filter, short exposure imaging and delayed acquisition strategy, which can effectively suppress the infrared background radiation of 1200℃ high-temperature forgings, significantly improve the signal-to-noise ratio of point cloud data, and realize high-quality, non-contact online data acquisition of hot forgings during the forging process, without waiting for the forgings to cool down, thus eliminating detection lag.
[0017] By fixing the feature surface of the mold with a constant position during the forging process, and combining it with the feature point cloud of the high-temperature resistant calibration block on site, registration is completed with the reference model. This achieves coordinate system unification for multiple process nodes and multiple acquisition devices, eliminates coordinate offset caused by station switching and equipment vibration, ensures the consistency of the reference for three-dimensional reconstruction and deviation calculation, and improves detection accuracy.
[0018] It enables tracking and monitoring of the entire forging process, covering all process nodes before, after, and after final forging. It can acquire the three-dimensional morphological changes of forgings in real time at each forming stage. It can not only determine the quality of the final forged product, but also detect the excessive deformation trend in the pre-forging stage in advance, providing data support for real-time adjustment of process parameters and reducing scrap generation from the source.
[0019] A multi-dimensional comprehensive deviation judgment system was constructed, integrating four core indicators: maximum local deviation, standard deviation of deviation distribution, volume deviation, and proportion of defect points. The comprehensive evaluation index is obtained by quantifying through a comprehensive judgment function. At the same time, the system combines local morphological deviation and curvature index to accurately identify defect areas, avoiding misjudgment and missed detection caused by a single size index.
[0020] It has the ability to adaptively update the threshold. Based on the statistical distribution characteristics of historical qualified samples, the judgment threshold is dynamically updated through an iterative formula. It can adapt to the normal fluctuations of raw material performance and process parameters during the forging production process, avoid over-inspection or under-inspection caused by fixed thresholds, and greatly improve the robustness and on-site adaptability of the detection system. It can be stably applied to industrial continuous forging production lines for a long time. Attached Figure Description
[0021] Figure 1 This is a flowchart illustrating the online reconstruction and deviation detection method for the three-dimensional morphology of forgings according to the present invention. Figure 2 This is a schematic diagram of the module of the online reconstruction and deviation detection system for the three-dimensional morphology of forgings of the present invention. Detailed Implementation
[0022] The present invention will now be described in further detail with reference to the embodiments and accompanying drawings. However, this should not be construed as limiting the scope of the above-described subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.
[0023] refer to Figure 1 Example 1: This example provides a method for online reconstruction and deviation detection of the three-dimensional morphology of forgings. The specific detection method includes the following steps: S1. Collect point cloud data of the forging at multiple process nodes before pre-forging, after pre-forging, and after final forging.
[0024] In this embodiment, three sets of line laser acquisition units are deployed at the pre-forging loading station, the pre-forging unloading station, and the final forging unloading station. Each acquisition unit includes a 405nm narrowband blue laser, a 5-megapixel global shutter industrial camera equipped with a 405nm±5nm narrowband filter, and an acquisition control industrial computer.
[0025] The acquisition process employs a high-temperature interference resistance strategy, specifically: a 405nm narrowband blue laser is emitted as the active light source and vertically irradiates the surface of the hot forging; the industrial camera filters out the infrared radiation emitted by the high-temperature forging through a narrowband filter, receiving only the laser reflection signal of the corresponding wavelength; the camera exposure time is set to 50μs, and acquisition is initiated after a 200ms delay after the robot places the forging at the inspection station and it comes to a complete stop. The short exposure suppresses the high-temperature background light, and the delayed acquisition eliminates the motion blur caused by the vibration of the forging, ultimately obtaining the original point cloud data of the forging from various process nodes and multiple perspectives.
[0026] S2. Obtain the fixed feature surface of the mold and the feature point cloud of the high-temperature resistant calibration block on site, and register the feature point cloud with the preset reference model to establish a globally unified fixed coordinate system.
[0027] In this embodiment, the mold is a crankshaft pre-forging mold and a final forging mold. The fixed feature surface of the mold is selected from the lower mold reference mounting surface and the end face of the four positioning holes. The position of this feature surface is fixed throughout the forging process and there is no relative displacement. The high-temperature resistant calibration block is made of silicon carbide material and has a temperature resistance of 1300℃. The calibration block is equipped with three feature balls of standard diameter and four positioning feature holes as the reference for global calibration.
[0028] The coordinate system establishment process is as follows: a three-dimensional reference CAD model of the mold and a standard model of the calibration block are pre-constructed; after the mold change is completed on the production line, the point cloud data of the fixed feature surface of the mold is collected and coarsely registered with the mold reference CAD model and finely registered with ICP to obtain the mold reference coordinate system; then the feature point cloud of the high-temperature resistant calibration block is collected and registered with the standard model of the calibration block to complete the hand-eye calibration, and the camera coordinate system of the acquisition unit of all stations is unified under the mold reference coordinate system to establish a globally unified fixed coordinate system, eliminate the coordinate deviation of multiple stations and multiple cameras, and ensure that the detection data reference of all process nodes is consistent.
[0029] S3. In the fixed coordinate system, the point cloud data is denoised and rigid body registered to obtain a three-dimensional reconstruction model of the forging.
[0030] The specific processing steps are as follows: First, the original point cloud data is denoised. A statistical filtering algorithm is used to calculate the average distance between each point and its 20 neighboring points. Outlier noise points with a distance exceeding three times the standard deviation of the mean are removed, as well as invalid points caused by environmental interference. Then, a bilateral filtering algorithm is used to smooth the denoised point cloud, which retains the key geometric features of the forging, such as the edges and corners, while smoothing the noise.
[0031] Subsequently, for the denoised point clouds from multiple perspectives under the same process node, the ICP iterative nearest point algorithm is used for rigid body registration and stitching. All point clouds are unified under the global fixed coordinate system established in step S2, and the complete watertight 3D reconstruction model of the forging is obtained by stitching. At the same time, the actual volume of the forging is calculated.
[0032] S4. Spatially align the three-dimensional reconstruction model with the preset standard model to obtain local shape deviation and volume deviation.
[0033] In this embodiment, the preset standard model is the standard CAD model of the crankshaft forging. The spatial alignment process is as follows: first, feature matching is performed through key feature points such as the main journal and connecting rod journal of the forging to complete the coarse registration of the three-dimensional reconstruction model and the standard model; then, the ICP algorithm is used for fine registration to achieve accurate spatial alignment of the two models, with the alignment accuracy controlled within 0.05mm.
[0034] The process of obtaining local topographic deviation is as follows: By calculating the directed distance, the Euclidean distance between each point in the current point cloud and its nearest corresponding point in the standard model is solved, yielding the local topographic deviation value Δ for each point. For each point, its 16 neighboring points are selected, and the normal vector of that point is calculated using principal component analysis. Then, the change in the angle between the normal vector of that point and the normal vectors of its neighboring points is calculated to obtain the local curvature index. The deviation threshold is set to 0.5 mm, and the curvature threshold is set to 0.2 mm. -1 Select samples that simultaneously satisfy Δ>0.5mm and curvature index>0.2mm. -1 Points that are not found are marked as defect points. The connected region formed by all defect points is the defect region. At the same time, the number of defect points n and the total number of points N in the point cloud are obtained.
[0035] The process of obtaining the volume deviation is as follows: the actual volume V of the forging is calculated through the three-dimensional reconstruction model, the theoretical volume V0 of the standard model is read, and the absolute value of the volume deviation |V-V0| is calculated; in this embodiment, the characteristic dimension D of the forging is selected as the total length of the crankshaft, which is used for the normalization of the volume deviation.
[0036] S5. Based on the local morphological deviation and volume deviation, a comprehensive evaluation index is calculated using a comprehensive judgment function.
[0037] In this embodiment, the expression for the comprehensive decision function is: ; Where J is the comprehensive judgment index, For the maximum local deviation, Here, V represents the standard deviation of the deviation, and V represents the current volume. Where D is the theoretical volume and D is the characteristic dimension of the forging. α = 0.4, β = 0.3, γ = 0.2, δ = 0.1. α, β, γ, and δ can be flexibly adjusted according to the precision requirements of different forgings.
[0038] Substitute the parameters obtained in step S4 into the above formula to calculate the comprehensive evaluation index J of the current forging.
[0039] S6. Compare the comprehensive evaluation index with the set threshold. When the set threshold is exceeded, it is determined that the forging deformation is excessive and an alarm signal is output.
[0040] In this embodiment, the initial threshold T0 is set to 0.8mm. The calculated comprehensive evaluation index J is compared with the set threshold T. If J>T, the forging is determined to be excessively deformed and has forming defects. An alarm signal is output through the sound and light alarm device, and at the same time, the non-conforming signal is sent to the production line PLC control system, triggering the robot to sort the forging to the non-conforming product area. If J≤T, the forging is determined to be qualified and flows into the next process.
[0041] Furthermore, this method also includes threshold adaptive updating, specifically: after inspecting every 50 qualified forgings, the mean μ and standard deviation σ of the comprehensive evaluation index of these 50 qualified samples are statistically analyzed, and the threshold is updated using the threshold adaptive updating formula, which is: ; The update step size η=0.2, confidence coefficient k=2, corresponding to a 95% confidence interval, T old T is the current threshold before the update. new This is the updated threshold. This adaptive update strategy can adapt to process fluctuations caused by raw material properties and mold wear, avoiding over- or under-detection issues resulting from a fixed threshold.
[0042] refer to Figure 2 Example 2: This example provides an online reconstruction and deviation detection system for the three-dimensional morphology of forgings, used to implement the detection method described in Example 1, such as... Figure 2 As shown, the system includes a data acquisition module, a coordinate establishment module, a reconstruction processing module, a deviation calculation module, an evaluation and judgment module, an alarm module, and a threshold update module.
[0043] The data acquisition module, deployed at various process inspection stations on the forging production line and linked to the production line control system, is used to collect point cloud data of the forging at multiple process nodes before, after, and after final forging when the robotic arm transports the forging to the inspection station. Specifically, the data acquisition module includes multiple sets of linear laser acquisition units. Each unit contains a 405nm-450nm narrowband blue laser, an industrial camera with a matching narrowband filter, an image acquisition card, and an acquisition control unit. Specifically, it is used to: illuminate the forging surface using a narrowband blue laser with a wavelength range of 405nm-450nm as an active light source; receive the laser reflection signal using an industrial camera equipped with a narrowband filter matching the laser wavelength; and suppress high-temperature background radiation through short-exposure imaging and delayed acquisition strategies to obtain high-quality point cloud data.
[0044] The coordinate establishment module communicates with the data acquisition module to acquire the fixed feature surfaces of the mold and the feature point cloud of the high-temperature resistant calibration block on site. It then registers the feature point cloud with a preset reference model to complete hand-eye calibration, establish a globally unified fixed coordinate system, and provide a unified coordinate reference for the entire process of inspection.
[0045] The reconstruction processing module is connected to the data acquisition module and the coordinate establishment module respectively. It is used to perform noise reduction and smoothing preprocessing on the acquired raw point cloud data under the fixed coordinate system, and then complete the multi-view point cloud stitching through rigid body registration to obtain a complete three-dimensional reconstruction model of the forging.
[0046] The deviation calculation module, which communicates with the reconstruction processing module, has a built-in standard model of the forging. It is used to spatially align the 3D reconstructed model with the preset standard model and calculate the local topographic deviation and volume deviation. Specifically, the deviation calculation module is used to: obtain the local topographic deviation value by calculating the Euclidean distance between each point in the current point cloud and the corresponding point in the standard model; obtain the local curvature index by calculating the change in the normal vector between the point and its neighboring points; and, based on the local topographic deviation value and the local curvature index, select points that simultaneously meet the deviation threshold condition and the curvature threshold condition to obtain the defect area, while simultaneously calculating the volume deviation.
[0047] The evaluation and judgment module is connected to the deviation calculation module and has a built-in preset comprehensive judgment function. It is used to calculate the comprehensive evaluation index based on the local morphological deviation and volume deviation. The expression of the comprehensive judgment function is shown in Example 1. The module can flexibly configure various weight coefficients to adapt to the detection requirements of different types of forgings.
[0048] The alarm module is connected in communication with the evaluation and judgment module and the production line control system. It is used to compare the comprehensive evaluation index with the set threshold. When the set threshold is exceeded, it is determined that the forging is deformed too much and an audible and visual alarm signal is output. At the same time, a non-conforming product sorting signal is sent to the production line control system.
[0049] The threshold update module communicates with the evaluation and judgment module and the alarm module respectively. It is used to obtain the sample distribution characteristics by statistically analyzing the mean and standard deviation of the deviation index of historical qualified samples, and to complete the iterative update of the threshold through a preset adaptive update formula. The updated threshold is synchronized to the alarm module for subsequent qualification judgment of forgings.
[0050] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods.
[0051] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.
Claims
1. A method for online reconstruction and deviation detection of the three-dimensional morphology of forgings, characterized in that, Includes the following steps: Point cloud data of the forging was collected at multiple process nodes before pre-forging, after pre-forging, and after final forging. Obtain the fixed feature surfaces of the mold and the feature point cloud of the high-temperature resistant calibration block on site, and register the feature point cloud with the preset reference model to establish a globally unified fixed coordinate system; In the fixed coordinate system, the point cloud data is denoised and rigid body registered to obtain a three-dimensional reconstruction model of the forging. The three-dimensional reconstruction model is spatially aligned with a preset standard model to obtain local shape deviations and volume deviations. Based on the local morphological deviations and volume deviations, a comprehensive evaluation index is calculated using a comprehensive judgment function. The comprehensive evaluation index is compared with the set threshold. When the set threshold is exceeded, it is determined that the forging is deformed excessively and an alarm signal is output.
2. The method for online reconstruction and deviation detection of three-dimensional morphology of forgings according to claim 1, characterized in that, The mold is a fixing device used to form the forging in the forging production line, and the fixing feature surface of the mold is a geometric feature surface on the mold whose position remains unchanged during the forging process; The point cloud data is collected using a high-temperature interference resistance strategy, specifically including: The surface of the forging is irradiated by emitting a narrow-band blue laser with a wavelength range of 405nm-450nm as an active light source. The laser reflection signal is received by an industrial camera equipped with a narrowband filter; By employing short-exposure imaging and delayed acquisition strategies, high-temperature background radiation is suppressed, resulting in high-quality point cloud data.
3. The method for online reconstruction and deviation detection of three-dimensional morphology of forgings according to claim 1, characterized in that, The obtained local morphological deviations specifically include: The local topography deviation value is obtained by calculating the Euclidean distance between each point in the current point cloud and the corresponding point in the standard model; The local curvature index is obtained by calculating the change in the normal vector between a point and its neighboring points. Based on the local morphological deviation value and the local curvature index, points that simultaneously meet the deviation threshold condition and the curvature threshold condition are selected to obtain the defect area.
4. The method for online reconstruction and deviation detection of three-dimensional morphology of forgings according to claim 3, characterized in that, The expression obtained through the comprehensive decision function: ； Where J is the comprehensive judgment index, For the maximum local deviation, Here, V represents the standard deviation of the deviation, and V represents the current volume. Where D is the theoretical volume and D is the characteristic dimension of the forging. Where N is the number of defect points, and N is the total number of points. , , , These are the weighting coefficients.
5. The method for online reconstruction and deviation detection of three-dimensional morphology of forgings according to claim 1, characterized in that, It also includes threshold adaptive updates, specifically including: The sample distribution characteristics are obtained by statistically analyzing the mean and standard deviation of the deviation indicators of historical qualified samples. The expression for threshold adaptive update is: ； Among them, T new The updated threshold To update the step size, T old The old threshold =2, This represents the mean of the deviation index.
6. A system for online reconstruction and deviation detection of three-dimensional morphology of forgings, characterized in that, include: The data acquisition module is used to collect point cloud data of the forging at multiple process nodes before pre-forging, after pre-forging, and after final forging. The coordinate establishment module is used to acquire the fixed feature surfaces of the mold and the feature point cloud of the high-temperature resistant calibration block on site, and to register the feature point cloud with the preset reference model to establish a globally unified fixed coordinate system. The reconstruction processing module is used to denoise and rigidly register the point cloud data in the fixed coordinate system to obtain a three-dimensional reconstruction model of the forging. The deviation calculation module is used to spatially align the three-dimensional reconstructed model with a preset standard model to obtain local topographic deviations and volume deviations. The evaluation and judgment module is used to calculate the comprehensive evaluation index based on the local morphological deviation and volume deviation through a comprehensive judgment function. The alarm module is used to compare the comprehensive evaluation index with the set threshold. When the set threshold is exceeded, it is determined that the forging deformation is excessive and an alarm signal is output.
7. The method for online reconstruction and deviation detection of three-dimensional morphology of forgings according to claim 6, characterized in that, The data acquisition module is specifically used for: The surface of the forging is irradiated by emitting a narrow-band blue laser with a wavelength range of 405nm-450nm as an active light source. The laser reflection signal is received by an industrial camera equipped with a narrowband filter; High-quality point cloud data was obtained by suppressing high-temperature background radiation through short-exposure imaging and delayed acquisition strategies.
8. The method for online reconstruction and deviation detection of three-dimensional morphology of forgings according to claim 6, characterized in that, The deviation calculation module is specifically used for: The local topography deviation value is obtained by calculating the Euclidean distance between each point in the current point cloud and the corresponding point in the standard model; The local curvature index is obtained by calculating the change in the normal vector between a point and its neighboring points. Based on the local morphological deviation value and the local curvature index, points that simultaneously meet the deviation threshold condition and the curvature threshold condition are selected to obtain the defect area.
9. The method for online reconstruction and deviation detection of three-dimensional morphology of forgings according to claim 8, characterized in that, In the evaluation and judgment module, the expression of the comprehensive judgment function is: ； Where J is the comprehensive judgment index, For the maximum local deviation, Here, V represents the standard deviation of the deviation, and V represents the current volume. Where D is the theoretical volume and D is the characteristic dimension of the forging. Where N is the number of defect points, and N is the total number of points. , , , These are the weighting coefficients.
10. The method for online reconstruction and deviation detection of three-dimensional morphology of forgings according to claim 6, characterized in that, It also includes a threshold update module, which is used to obtain the sample distribution characteristics by statistically analyzing the mean and standard deviation of the deviation indicators of historical qualified samples, and then adaptively update the threshold. The expression for the adaptive threshold update is: ; Among them, T new The updated threshold To update the step size, T old The old threshold =2, This represents the mean of the deviation index.