Papers by Keyword: Process Monitoring

Abstract: In forging, parallelism deviations between upper and lower dies lead to asymmetric flash formation and affect forming forces, die filling, and part integrity. The flash gap influences local flow resistance and is closely linked to flow behavior and dimensional precision. Conventional diagnostics often assess such deviations under no-load or quasi-static conditions and therefore may not capture the effective closing state at bottom dead centre (BDC) under process load. While modern approaches such as high-resolution optical tracking of ram deflection can provide valuable insight, they require dedicated and sensitive instrumentation and are often limited in scalability. In contrast, workpiece-based signatures inherently reflect process effects such as elastic deflections, guide clearances, frictional conditions, and thermal influences.This study investigates whether workpiece-related geometric features can serve as diagnostic signatures for detecting and quantifying closing-gap inclinations under load. The focus is on the locally resolved flash thickness, which reflects the effective closing gap at BDC. Because this gap results from both geometric alignment and load-dependent deformation, the evaluation targets the final load-bearing state. Comparative forging trials are performed on a press equipped with active parallelism control, where controlled misalignments are introduced. The resulting flash geometry is measured by laser triangulation to determine the resolution limit and to identify the deviation magnitude at which reproducible signatures can be detected under process-relevant conditions. In the investigated setup, flash-thickness asymmetry shows an increasing trend from closing-gap inclinations of ~0.25°, providing a markedly higher diagnostic sensitivity than the maximum forming force. Designed as a non-invasive and retrofit-capable method, the approach supports inline monitoring in high-volume forging. It further enables scalable, data-driven correlation of machine, process, and product data for condition-aware process optimization.

Abstract: A novel solution for monitoring the infusion process and providing decision support to operators involved in the manufacturing of large, unique or near-unique parts is presented. Based on a scientific approach referred to as the 5D methodology (D for dimensions), the proposed solution consists of a process digital twin built upon a metamodel that is fed in real time by signals from sensors embedded in the process, enabling the anticipation of defects such as dry spots.

Abstract: Bobbin-Tool Friction Stir Welding (BT-FSW) is a solid-state joining process in which axial forces are internally balanced by the tool, eliminating the need for a backing plate and enabling the joining of hollow aerospace structures. Owing to the coupled thermo-mechanical nature of the process, weld stability and quality are governed by the interaction between process parameters and the resulting torque response, which is difficult to assess in situ using conventional sensing alone. BT-FSW experiments were performed on AA2024-T351 sheets with thicknesses of 2.4, 2.8 and 3.6 mm using a structured Design of Experiments (DoE). The 3.6 mm joints achieved approximately 90 % of the base-material strength, while the 2.4 and 2.8 mm joints reached about 80 % and 85 %, respectively. These mechanical results were used as ground truth to train machine learning regression models for steady-state torque predictions. By augmenting nominal process parameters with force-derived features, the proposed soft-sensing framework achieved strong agreement between predicted and measured torque, demonstrating that compact, physics-based feature engineering enables reliable prediction under limited experimental data conditions.

Abstract: In modern precision manufacturing, optimizing complex processes like turn-milling is crucial for reducing production costs and ensuring high surface integrity. In this study the application of artificial intelligence, specifically machine learning (ML), for modeling turn-milling processes is investigated. The complexity of machining operations and the multitude of influencing input parameters often lead to time-consuming setups, particularly in single-part or small series manufacturing. Traditional process monitoring methods frequently fall short due to system complexity, prompting the exploration of ML for process optimization and automation. Focusing on orthogonal turn-milling, experimental data was collected to address regression problems such as tool wear and surface roughness, as well as tool condition classification. Three regression models - linear, polynomial, and support vector regression (SVR) - and four classification models - logistic regression, neural networks, support vector machines (SVM), and decision trees - were trained and validated using k-fold cross-validation. For regression models, root mean square error (RMSE) was used as the performance evaluation metric, while accuracy and F1-score were employed for classification problems. The results indicate that ML algorithms provide enhanced flexibility and accuracy compared to traditional statistical techniques, offering potential reductions in time and costs in process setups. By optimizing parameters iteratively, ML models demonstrate higher precision, reducing the need for extensive empirical research and the associated experimental costs. The developed models can be adapted to various manufacturing processes with minimal code adjustments, broadening their applicability and efficiency.

Abstract: The Extrusion Benchmark 2023 was focused on the evaluation of different die design strategies for the manufacturing of AA6082 hollow tubes (40 mm external diameter and 4 mm thickness) through a porthole die with 3 openings. The extrusion process was monitored in industrial environment in terms of press load, profiles speed, profiles exit temperature, and die temperatures under different processing conditions (air quenching, water quenching, nitrogen die cooling). Extruded profiles were then analyzed in terms of seam weld quality, charge weld extension and microstructure evolution for both air/water quench and presence/absence of nitrogen cooling. The results of the study are aimed at validating FEM simulation outputs in the context of the International Conference on Extrusion and Benchmark (ICEB).

Abstract: Laser metal deposition (LMD) industrial use demands research about the influence of the parameters in the built parts density, accuracy and mechanical properties. Especially for the thin-wall parts, knowledge about the correlations between processing parameters and the final result is indispensable. This study explores the relationship between process parameters and the quality of AISI316L stainless steel thin-walled parts produced by LMD. A six-axis robot equipped with a deposition head allowed relative spatial movement between the powder nozzle and laser beam and substrate with high accuracy. Controlled energy input provided by continuous wave Ytterbium fibre laser allows using less material flow rate and the production of thin layers in test samples. Three processing parameters were selected to investigate the effects on part characteristics using a Box-Behnken experimental design. Through this method, each parameter was evaluated between 600 W to 800 W laser power, 6 mm/s to 14 mm/s feedrate and 0.2 mm to 0.4 mm layer thickness. All remaining parameters were fixed using argon to provide an inert atmosphere, 8.8 g/min powder feeding rate and 1.5 mm spot diameter. The method was used to test the manufacture of thin-wall cylindrical specimens with 10 mm in height and 75 mm in diameter. Fabricated AISI316L samples were evaluated regarding the dimensional and geometrical characteristics. It was observed that higher energy input density during the laser additive manufacturing implies lower geometric precision. Feedrate and layer thickness has the highest impact on both the wall thickness and vertical accuracy. Given the inability to produce parts with an acceptable final surface, the process finds great applicability when complemented with additional finishing technologies.

Abstract: Clinching is a cost efficient method for joining components in series production. To assure the clinch point’s quality, the force displacement curve during clinching or the bottom thickness are monitored. The most significant geometrical characteristics of the clinch point, neck thickness and undercut, are usually tested destructively by microsectioning. However, micrograph preparation goes ahead with a resetting of elastic deformations and crack-closing after unloading. To generate a comprehensive knowledge of the clinch point’s inner geometry under load, in-situ computed tomography (CT) and acoustic testing (TDA) can be combined. While the TDA is highly sensitive to the inner state of the clinch point, it could detect critical events like crack development during loading. If such events are indicated, the loading process is stopped and a stepped in-situ CT of the following crack and deformation development is performed. In this paper, the concept is applied to the process of clinching itself, providing a detailed three-dimensional insight in the development of the joining zone. A test set-up is used which allows a stepwise clinching of two aluminium sheets EN AW 6014. Furthermore, this set-up is positioned within a CT system. In order to minimize X-ray absorption, a beryllium cylinder is used within the set-up frame and clinching tools are made from Si3N4. The actuator and sensor necessary for the TDA are integrated in the set-up. In regular process steps, the clinching process is interrupted in order to perform a TDA and a CT scan. In order to enhance the visibility of the interface, a thin tin layer is positioned between the sheets prior clinching. It is shown, that the test-set up allows a monitoring of the dynamic behaviour of the specimen during clinching while the CT scans visualize the inner geometry and material flow non-destructively.

Abstract: Process monitoring in additive manufacturing (AM), i.e. in laser powder bed fusion (LPBF) of metal parts, has been identified as the crucial bottleneck in accelerating the AM industrialization process. To reduce the cost and time needed to produce and qualify an AM part, an online monitoring system of the manufacturing process is desirable. While the currently available systems capture a large amount of process data, they still lack the ability to interpret the acquired data adequately. In this work we present the first steps towards an automated evaluation of online monitoring data i.e. melt pool data. It is shown that a well-trained convolutional neural network (CNN) is able to detect artificially induced process deviations on the basis of melt pool characteristics.

Abstract: Laser powder bed fusion (LPBF) is a frequently used manufacturing process due to its advantages in lightweight construction, design possibilities and functionalization of geometry. However, the printed parts will often have to undergo time and cost expensive non-destructive testing by sophisticated methods like X-CT. Thus, there is a strong demand to identify suitable online process monitoring techniques that allow to reduce or substitute post-process NDT effort. The temperature field reacts sensitively to deviations during processing, thus online temperature monitoring is a promising approach. In the present work a spatially resolved temperature measurement, based on 2-channel-pyrometry, is used for process monitoring in LPBF. The camera system is coaxially integrated into the beam guidance of the LPBF system. The coaxial observation enables a lateral resolution better than 10 μm over the whole build-up area of 250 x 250 mm2. Single tracks were welded with different parameters and observed by the camera system to identify thermal indicators. Metallographic cross-sections of the tracks were compared with the melt pool width measured by the online observation system. The deviation was ca. 3 %. In addition, cubes of 10 mm by 10 mm by 10 mm are built up. The melt pool area is identified as useful indicator for the process behavior and for the first time the assessment of part density is demonstrated in LPBF during process by the help of a thermal monitoring system.

Abstract: Understanding and predicting relationships between laser welding process parameters, such as laser power and welding speed, and molten pool have been studied widely in order to critically control and improve laser welding. The laser welding processes are difficult to monitor in real time because of high temperature and rapid heating characteristics. In this study, infrared camera was set to collect data and provide real time monitoring system to determine the molten pool characteristics and weld quality. This study carried out a laser welding of SS400 low carbon steel and analyzed real-time image of the welding process to determine the average temperature of molten pool and calculate the size of molten pool. By varying the laser power and the welding speed, the infrared camera and imaging processing technique can monitor change of molten pool temperature in a range of 1000 C to 15000 C with about 1% temperature fluctuation. In addition, the size of molten pool can be calculated from the temperature profile of the welding zone. The calculated molten pool size was about 95% accurate compared to the measured size from microscope imaging.