Monitoring unit for laser cladding of metallic components and methods for monitoring laser cladding.

Abstract

The invention relates to a monitoring unit for laser cladding of metallic components, in particular for surface coating and / or structural build-up using a metallic and / or ceramic powder, comprising a laser beam source (1), a powder feed device (2), a welding nozzle (3), a spatially resolved camera (4), and an image data processing unit (5), wherein the image data processing unit (5) is configured to automatically evaluate the image data of the laser-clad and coated surface captured by the camera (4) with regard to the coating quality during or after the welding process using artificial intelligence.

Description

[0001] The present invention relates to a monitoring unit for laser cladding of metallic components, in particular for surface coating and / or structural fabrication using a metallic and / or ceramic powder, comprising a laser beam source, a powder feed device, a welding nozzle, a spatially resolved camera, and an image data processing unit. The invention further relates to a method for monitoring the laser cladding process.

[0002] In the context of laser metal deposition (LMD), quality assurance is of paramount importance for the surface coating and structural build-up of metallic components. Nevertheless, significant disadvantages and problems exist in established practice, which impair the efficiency and precision of the manufacturing process.

[0003] Currently, destructive methods, such as cross-sectioning, are primarily used to assess coating quality. This process is not only material- and time-consuming, but also means that the component itself can no longer be used. This leads to increased costs and material consumption, as a portion of the manufactured component must be sacrificed for each quality check.

[0004] Another characteristic used to assess quality is surface roughness, but measuring it requires additional time and an extra process step. This not only slows down the overall production process but can also be a source of further costs.

[0005] Furthermore, subjective visual inspection is a common method, although it is prone to errors and time-consuming. Human assessment of coating quality can be inconsistent and vary considerably depending on the operator's experience, thus reducing the reliability and reproducibility of quality control.

[0006] The accuracy of current quality assurance measures is subject to significant fluctuations and is also characterized by a low degree of automation. This means that numerous processes are carried out manually rather than automatically, which does not satisfactorily meet modern requirements for efficiency, cost savings, and consistently high product quality.

[0007] Finally, while imaging techniques like VisionLine are already used in industrial practice, they are primarily employed for positioning tasks and not as an integral part of quality assurance. A lack of intelligent integration of these systems into the coating quality assessment process represents an untapped potential for process optimization and quality improvement.

[0008] The object of the invention is therefore to provide a solution for the complex and error-prone process of evaluating coating quality in laser cladding, which currently leads to a high reject rate. Furthermore, the invention aims to provide a means of automatically detecting coating quality limitations and informing the machine operator or production manager of any existing quality deficiencies. Another object of the invention is to provide a device capable of identifying the cause of detected quality defects and suggesting possible solutions based on this identification. Finally, the invention aims to develop a monitoring unit for laser cladding that enables non-destructive testing while requiring only minimal additional time, as the testing is to be performed directly at the welding system.Another objective of the invention is to replace the shortcomings of a subjective assessment with unambiguous and low-error image data processing in order to realize a more reliable and accurate quality assessment system.

[0009] This task is solved by a monitoring unit for laser cladding of metallic components, in particular for surface coating and / or structural build-up using a metallic and / or ceramic powder, comprising a laser beam source, a powder feed device, a welding nozzle, a spatially resolved camera, and an image data processing unit, wherein the image data processing unit is configured to automatically evaluate the image data of the laser-clad and coated surface captured by the camera with regard to the coating quality during or after the welding process using artificial intelligence.

[0010] The monitoring unit according to the invention offers the advantage that the quality of laser-deposited and coated surfaces can be reliably assessed without human intervention. The integration of artificial intelligence enables an objective and consistent evaluation of the coating quality, thereby reducing defect rates and increasing process reliability. Furthermore, the automation of the evaluation process allows for significant time and cost savings, as manual, non-destructive testing is no longer required.

[0011] The monitoring unit, equipped with an image data processing unit that uses artificial intelligence to evaluate the data captured by a camera of the laser-clad and coated surface, offers further advantages in addition to reliable quality assessment. The automated evaluation of coating quality relieves the operator of subjective assessments and tedious inspection tasks, thus increasing the efficiency and accuracy of quality assurance.

[0012] The use of the system results in a clear OK / not OK (OK / NOO) classification of the determined coating quality, which leads to clear decision-making and thus reduces the reject rate and increases product reliability.

[0013] Integrating this unit into the welding process enables non-destructive testing, eliminating the need for additional post-weld inspections and significantly reducing the time required for subsequent quality checks. Image data processing ensures a clear and error-free evaluation, eliminating the drawbacks of the subjective, visual assessments previously common. The detection capabilities of artificial intelligence, in combination with advanced sensors such as the Visionline camera, further enhance the reliability of the inspection, as the machine can detect even the finest defects that would remain invisible to the human eye.

[0014] The use of artificial intelligence (AI) for image data processing offers significant advantages in the monitoring and evaluation of laser-deposited surface structures. A key advantage is that surface structure images typically lack easily identifiable features, quickly pushing conventional image processing methods to their limits. The surface finish of a laser-deposited component is often complex and dependent on a multitude of different influencing factors, making the detection and evaluation of these features difficult without assistance.

[0015] Integrating AI methods into the image data processing process effectively manages this high level of complexity. AI algorithms are capable of reliably identifying and interpreting even subtle and not immediately obvious structural elements. These advanced algorithms learn from a wide variety of data patterns and continuously improve to enable increasingly accurate analysis.

[0016] Another key advantage is AI's ability to autonomously recognize relevant features and assess their significance. Thanks to this self-learning capability, AI can identify and classify crucial aspects, such as the weld seam's condition, potential defects, or surface anomalies. This enables a more objective and accurate evaluation of weld quality than could be achieved by human observers. Ultimately, this quality assessment contributes to greater product consistency and reliability, thereby minimizing the risk of production errors and scrap.

[0017] The present invention relates to a monitoring unit and a method for laser cladding of metallic components, which can be used for both surface coating and structural fabrication. The laser cladding process uses metallic and / or ceramic powders, typically iron-based materials. Alternatively, nickel-, cobalt-, or titanium-based materials, or other materials, can also be used.

[0018] The process typically involves applying one or more layers, with each layer being individually evaluated. The layer thicknesses of these applications typically range from 1 to 1000 micrometers, with thicknesses between 50 and 500 micrometers and especially between 80 and 200 micrometers per layer being particularly common.

[0019] Furthermore, the coating materials used are characterized by grain sizes of 15 to 120 micrometers, with grain sizes of 15 to 45 micrometers being particularly common for high-speed cladding.

[0020] The carrier gas for the coating material is typically an inert gas such as nitrogen, argon, helium, CO2, or a mixture of these gases. Depending on the application, the laser beam source can be a single-mode laser, such as a TruFiber 500-2000, or a multi-mode disk laser, for example, a TruDisk 2000-6000. Laser beam sources such as the TruDisk 6001-10001 or TruDisk 16002 are preferred. The laser power required for laser cladding varies depending on the desired welding result and can range from 50 watts to 2000 watts for structural build-up or from 1.5 kilowatts to 24 kilowatts for area coating. For optimal process control, a beam parameter product (SPP) of 100 millimeters by milliradians or less and a beam diameter of no more than 5 millimeters, but preferably between 2 and 3 millimeters, are used.Furthermore, the mass flow rate of the coating material is typically between 0.8 and 130 grams per minute. The laser beam source used can be either an infrared laser with a wavelength in the range of 800 to 1200 nanometers, preferably at 1030 or 1070 nanometers, or alternatively a VIS laser, in particular with wavelengths of 400 to 450 nanometers (blue) and 515 nanometers (green).

[0021] Another essential element of the monitoring unit is a spatially resolved camera, preferably with a CCD or CMOS sensor, which can be arranged coaxially or laterally / obliquely with respect to the laser beam; the coaxial arrangement being preferred. To protect the camera from interference emissions, such as back-reflected laser radiation or process lights, it is equipped with optical filters such as neutral density filters and / or bandpass filters.

[0022] The image data processing unit is essential for evaluating the data captured by the camera. This unit processes the image data using artificial intelligence (AI) to independently recognize and assess relevant features. Alternatively, feature detection based on grayscale image contrasts is possible, but without AI support.

[0023] Crucial to the process is that each individual layer is evaluated separately, as each layer consists of numerous superimposed individual tracks. The outermost, most recently welded track is preferentially used for quality assessment. This targeted evaluation enables precise and reliable quality control of the laser cladding process.

[0024] According to an advantageous embodiment of the invention, the camera can be arranged coaxially to the laser beam of the laser source. This ensures that the camera is always in focus and that the layers to be evaluated can be inspected directly. This positioning also facilitates the automated alignment and adjustment of the camera system when the welding process changes, thus contributing to increased accuracy and reliability of the quality assessment. A key advantage of this arrangement is that the camera's measuring range is always precisely within the area of ​​the last welded seam. This ensures that the most relevant data of the weld seam are captured without distortion or perspective deviation, thus improving the accuracy and reliability of the quality assessment. Another advantage of coaxial camera positioning is its ease of use.Because the camera and laser beam are on a common axis, no additional adjustments or alignment are required to optimize the recording angle. This saves time during installation and calibration and reduces the complexity of system setup. Furthermore, the coaxial camera arrangement reduces the need for additional hardware. Since the camera is often already an integral part of the welding system, no additional hardware is required to implement the monitoring system. This reduction in components means that both the cost and complexity of retrofitting or purchasing new equipment are reduced, making the monitoring unit more economical and user-friendly overall.

[0025] According to a further preferred embodiment of the invention, the camera for image data acquisition can also be mounted at an angle to the laser beam of the laser source to provide an improved perspective of the weld layer and any irregularities. This allows potential irregularities and defects that would not be visible from other angles to be detected and corrected. This improves quality assurance and reduces the risk of defects that could later lead to product failures or quality issues.

[0026] Furthermore, according to an advantageous embodiment of the invention, the camera can be equipped with optical filters, in particular neutral density filters and / or bandpass filters, to minimize interference emissions, back-reflected laser radiation, and / or process lights. This results in clearer and interference-free image data acquisition, thereby improving the accuracy of automated image evaluation. Consequently, coating quality can be better assessed and false evaluations reduced, thus increasing the reliability and precision of the monitoring unit.

[0027] According to a further particularly preferred embodiment of the invention, the camera may be provided with an illumination device for generating an illuminated field of view, in particular by means of fluorescent tubes, LEDs, or an illumination laser, which ensures uniform and intense illumination of the field of view. This makes features of the laser-deposited and coated surface more clearly visible and enables precise image acquisition. This contributes to an improved recognition rate in automated quality assessment and thus enables more efficient control.

[0028] Furthermore, the invention can also be further developed such that the camera is a CCD or CMOS sensor. The use of modern sensor technologies such as CCD or CMOS sensors increases image quality and processing speed. High-resolution sensors also detect minor surface deviations, which allows for more detailed analysis and consequently a more accurate assessment of coating quality. This contributes to effective and efficient production monitoring with lower error rates.

[0029] In a preferred embodiment of the invention, the laser beam source may also comprise an infrared laser with a wavelength in the range of 800 to 1200 nm. The use of an infrared laser with a wavelength in the range of 800 to 1200 nm offers the advantage of deeper penetration into metallic materials, which enables higher efficiency in the welding process. This allows for more precise control of the welding process and results in more uniform coatings, thus increasing the overall performance and quality level of the processing.

[0030] It can also be advantageous to further develop the invention such that the laser beam source has a power output in the range of 50 W to 24 kW, which allows for flexible adaptation to different materials and welding tasks. Thus, the monitoring unit can be used for a variety of metal types and coating thicknesses without the need for additional changeover times or specialized equipment. This contributes to the cost-effectiveness and versatility of the monitoring unit.

[0031] According to a further preferred embodiment of the invention, the image data processing unit can use the outermost, most recently welded single track of a layer to assess its quality. Evaluating the outermost, most recently welded single track of a layer increases the precision of the quality assessment because it provides the most up-to-date and therefore most meaningful information about the welding process. This enables the early detection and correction of defects before they become hidden in inner layers, leading to a significant reduction in scrap and optimized resource efficiency. Furthermore, this specified evaluation allows the entire width and depth of the layer track to be detected and analyzed in detail. This holistic view of the layer track ensures that any quality defects that may occur can be detected regardless of their position within the track.Another significant advantage of this approach is the ability to evaluate the bonding area in the lower part of the layer track. This area is highly relevant for layer adhesion and therefore crucial for the structural integrity of the component. Any defects or irregularities in the bonding area can significantly impair the component's stability, which can be detected and prevented through this type of evaluation. Furthermore, focusing on the last welded single track enables precise and continuous quality control in real time.

[0032] Finally, the object of the invention can also be achieved by a method for monitoring the laser cladding of metallic components using a monitoring unit according to any one of claims 1 to 9, wherein the image data of the welded and coated surface are captured by the camera and evaluated by the image data processing unit using artificial intelligence with regard to the coating quality. The AI-based monitoring method offers a systematic and data-driven quality assessment. Since the AI ​​learns independently and can adapt to new patterns, continuous improvement in the accuracy and reliability of the assessment is enabled over time. This contributes to an increase in production quality and a reduction in quality-related downtime and rework, which ultimately reduces production costs and increases customer satisfaction.

[0033] The invention will now be explained in more detail with reference to figures, without limiting the general concept of the invention.

[0034] It shows: Fig. 1 a monitoring unit for laser cladding of metallic components in a schematic representation, Fig. 2 a sigmoid activation function of AI training, Fig. 3. Temporal development of correct classifications during AI training.

[0035] The Fig. Figure 1 shows a monitoring unit for laser cladding of metallic components, in particular for surface coating and / or structural construction using a metallic and / or ceramic powder, with a laser beam source 1, a powder feed device 2, a welding nozzle 3, a spatially resolved camera 4, and an image data processing unit 5.

[0036] The image data processing unit 5 is set up to automatically evaluate the coating quality of the image data of the laser-clad and coated surface captured by the camera 4 during or after the welding process using artificial intelligence.

[0037] Camera 4 is arranged coaxially to the laser beam of laser beam source 1. As shown in the Fig. As shown in Figure 1, the camera 4 is equipped with optical filters 6, in particular gray filters and / or bandpass filters, to minimize interference emissions, back-reflected laser radiation and / or process lights.

[0038] The camera 4 is designed as a CCD or CMOS sensor and has a lighting device 7 for generating an illuminated field of view, in particular by means of fluorescent tubes, LEDs or an illumination laser.

[0039] The laser beam source 1 comprises an infrared laser with a wavelength in the range of 800 to 1200 nm and a power in the range of 50 W to 24 kW.

[0040] The method for monitoring the laser cladding of metallic components using a monitoring unit, as described in the Fig. As shown in Figure 1, the process can proceed as follows. First, the image data of the welded and coated surface is captured by camera 4 and evaluated by image data processing unit 5 using artificial intelligence with regard to coating quality. Image data processing unit 5 uses the outermost, most recently welded single track of each layer to assess its quality.

[0041] Training the AI ​​image evaluation for the monitoring unit in laser cladding is a multi-stage process that aims to develop a model that can effectively distinguish between high-quality (good) and defective (bad) weld layers.

[0042] Three sets of image data were used for training, each set consisting of 360 images. These images were captured coaxially through the nozzle hole, scanning the applied path at 1-degree intervals to obtain a wide variety of visual data representing the different possible states of the weld.

[0043] Based on the initially available dataset, an AI model was trained to reliably distinguish between high-quality images (H18) and low-quality images (H11). After training, this model was applied to a new, previously unseen category (H13), which also represents defective welding results.

[0044] The dataset was divided so that 60% of the images were used to train the model, 20% were used for validation to optimize and fine-tune the model, and the remaining 20% ​​were used to test the model and evaluate its performance. This corresponds to a distribution of 72 images per class for training, validation, and testing.

[0045] As a result of the AI ​​training, it was found that the sigmoid activation function is an effective method for classification into two classes. This is in the Fig. Figure 2 shows that, typically, a sigmoid function output value less than 0.5 is assigned the class "bad" and a value greater than 0.5 is assigned the class "good". With this classification, the model clearly distinguished between good and bad images in the dataset without any classification errors—there were no false positives or false negatives. It is also important to observe the temporal evolution of the correct classifications during training. The model achieved good performance on the validation dataset from time point 4 of the training process onward (see also Figure 2). Fig. 3) This indicates that from this point on, the model has learned to recognize and correctly classify relevant features, and confirms the effective adaptability of the AI ​​to the given quality characteristics.

[0046] In summary, the targeted collection and processing of image data, the methodology of data splitting, and the continuous monitoring and optimization system during the training of the AI ​​enable high reliability and accuracy of the trained model for automated quality monitoring in laser cladding.

[0047] The invention is not limited to the embodiments illustrated in the figures. The foregoing description is therefore not to be considered limiting, but rather explanatory. The following claims are to be understood as meaning that a named feature is present in at least one embodiment of the invention. This does not preclude the presence of further features. Insofar as the claims and the foregoing description define 'first' and 'second' features, this designation serves to distinguish between two similar features without establishing any hierarchy. Reference symbol list 1 Laser beam source 2 Powder feeder 3 welding nozzles 4 cameras 5 Image data processing unit 6 filters 7 Lighting equipment

Claims

[1] Monitoring unit for laser cladding of metallic components, in particular for surface coating and / or structural build-up using a metallic and / or ceramic powder, with a) a laser beam source (1), b) a powder feeding device (2), c) a welding nozzle (3), d) a spatially resolved camera (4), and e) an image data processing unit (5), characterized by , that the image data processing unit (5) is set up to automatically evaluate the image data of the laser-clad and coated surface captured by the camera (4) with regard to the coating quality during or after the welding process using artificial intelligence. [2] Monitoring unit according to claim 1, characterized by that the camera (4) is arranged coaxially to the laser beam of the laser beam source (1). [3] Monitoring unit according to claim 1, characterized by , that the camera (4) for image data acquisition is mounted in an oblique arrangement to the laser beam of the laser beam source (1) in order to provide an improved perspective on the weld layer and any irregularities. [4] Monitoring unit according to any of the preceding claims, characterized by , that the camera (4) is equipped with optical filters (6), in particular gray filters and / or bandpass filters, to minimize interference emissions, back-reflected laser radiation and / or process lights. [5] Monitoring unit according to any of the preceding claims, characterized by , that the camera (4) has a lighting device (7) for producing an illuminated field of view, in particular by means of fluorescent tubes, LEDs or an illumination laser. [6] Monitoring unit according to any of the preceding claims, characterized by , that the camera (4) is a CCD or CMOS sensor. [7] Monitoring unit according to any of the preceding claims, characterized by , that the laser beam source (1) comprises an infrared laser with a wavelength in the range of 800 to 1200 nm. [8] Monitoring unit according to any of the preceding claims, characterized by , that the laser beam source (1) has a power output in the range of 50 W to 24 kW [9] Monitoring unit according to any of the preceding claims, characterized by , that the image data processing unit (5) uses the outermost, last welded single track of a layer to assess the quality. [10] Method for monitoring the laser cladding of metallic components using a monitoring unit according to any one of claims 1 to 7, characterized by, that the image data of the welded and coated surface is captured by the camera (4) and evaluated by the image data processing unit (5) using artificial intelligence with regard to the coating quality.

Images