Using wearable sensors to evaluate workers’ performance is challenging with existing sensor techniques. It requires detecting not only limb motions but also the onset and offset of specific actions. Commonly used inertial measurement units (IMUs) can be combined with surface electromyography (sEMG) to detect muscular activity. However, sEMG requires skin preparation and careful sensor placement, and can be affected by sweat or motion artifacts. To address these limitations, we used a wearable system combining IMUs and force-sensing resistors (FSRs), where IMUs capture joint kinematics and FSRs detect grasping actions. The system included three IMUs (on the trunk, upper arm, and forearm) and two FSR arrays (on the upper and lower arms). The system was first validated in a laboratory setting against an optical motion capture system with 10 healthy young adults performing isolated upper limb movements and mimicking lifting tasks. The results showed high agreement in joint angle estimation (coefficient of multiple correlation = 0.95 $ \pm $ 0.04), with a maximum root mean square error of 8.7 $ \pm $ 2.92°, and a mean absolute timing error for grasp detection of −0.59 seconds. To evaluate its applicability in real-world scenarios, a pilot in-field test was then conducted with two manufacturing workers (using and not using a passive shoulder exoskeleton) during a repetitive panel-packing task. The test shows highly consistent grasping detection, which allowed segmenting the task with a small variability in task duration (maximum coefficient of variation = 5.16$ \% $). These findings demonstrate the feasibility of using the proposed method in industrial environments to analyze upper limb motion and grasping activity.

The role of pylon-induced vortex structures on flame stabilisation within a supersonic pylon-cavity flameholder is numerically investigated. The study examines how the fuel jet interacts with the vortices produced by three distinct pylon-cavity flameholder geometries labelled as P0, P1 and P2. P0 represents the pyramidal-shaped baseline pylon configuration, whereas P1 and P2 consist of parallel and slanted grooves on the pylon slant surfaces with respect to the supersonic crossflow for the generation of instream vortices. The selection criteria for P1 and P2 are based on reduced effective blockage area compared with P0. The inlet flow Mach number used for the investigation is 2.2. A sonic ${{\rm{H}}_2}$ fuel injection at 2.5 bar and 250 K is used for all the test cases. Steady RANS reactive flow simulations are used for the assessments. An 18-step Jachimowski chemical kinetic scheme is used to model ${{\rm{H}}_2}$-air reaction mechanism. The flowfield structures within the pylon-cavity flameholder are categorised into two, (i) pylon-cavity geometry-induced vortex structures (II, III, and IV) and (ii) fuel jet vortex pairs (FJVP and SFJVP). The study shows that the interaction between these two decides the reactant mixture formation within the flameholder and the flame stabilisation. This study identifies four different flame-holding locations – L1, L2, L3, and L4 – and their strength depends on the pylon configuration. Overall the P2 configuration is found to perform better than the others in terms of high heat release magnitude and flame spread within the combustor.

Despite their widespread use, purely data-driven methods often suffer from overfitting, lack of physical consistency, and high data dependency, particularly when physical constraints are not incorporated. This study introduces a novel data assimilation approach that integrates Graph Neural Networks (GNNs) with optimization techniques to enhance the accuracy of mean flow reconstruction, using Reynolds-averaged Navier–Stokes (RANS) equations as a baseline. The method leverages the adjoint approach, incorporating RANS-derived gradients as optimization terms during GNN training, ensuring that the learned model adheres to physical laws and maintains consistency. Additionally, the GNN framework is well-suited for handling unstructured data, which is common in the complex geometries encountered in computational fluid dynamics. The GNN is interfaced with the finite element method for numerical simulations, enabling accurate modeling in unstructured domains. We consider the reconstruction of mean flow past bluff bodies at low Reynolds numbers as a test case, addressing tasks such as sparse data recovery, denoising, and inpainting of missing flow data. The key strengths of the approach lie in its integration of physical constraints into the GNN training process, leading to accurate predictions with limited data, making it particularly valuable when data are scarce or corrupted. Results demonstrate significant improvements in the accuracy of mean flow reconstructions, even with limited training data, compared to analogous purely data-driven models.