Complex pipeline network systems widely distributed in industrial production are prone to damage due to long-term exposure to high-temperature and pressure environments. Traditional detection is based on manual periodic disassembly, which is costly and risky. The existing detection robots typically struggle to achieve in-situ detection of the entire pipe system. Therefore, we propose a dual-body support-adsorption wheeled rigid-soft hybrid robot that can detect each component in a pipeline network system, including pipes, valves, and containers. The robot adopts an adaptive variable-diameter support module. A vacuum adsorption module featuring a polyurethane sponge seal is utilized to ensure stable adhesion on walls. The posture adjustment of the robot in the complex pipeline system is realized by controlling a flexible skeleton. The results show that the robot can dexterously pass through complex structures, e.g., pipe elbows, stop valves, containers, and T-knots. It can be equipped with a camera pan, a miniature gripper, and other loads to achieve image acquisition and object capture. This study provides an efficient solution for in situ non-destructive detection of complex pipe systems in nuclear, chemical, and other industries.
A hybrid physics-informed neural network (PINN) combined with NSGA-II is presented in this work for the purpose of predicting and optimizing the tribological performance of direct metal laser sintered (DMLS) Inconel 718 under dry-sliding conditions. A pin-on-disc tribometer was used to collect experimental data while tracking a Taguchi L9 orthogonal array. The input parameters were sliding velocity (1–3 m/s), sliding distance (1000–2000 m), and applied load (10–30 N). The suggested PINN architecture, which incorporates Archard’s wear law straight into the loss function to ensure physical consistency, was used to model four output responses: wear rate (WR), specific wear rate (SWR), frictional force (FF), and coefficient of friction (CoF). With R2 = 0.900 (WR), 0.901 (SWR), 0.985 (FF), and 0.856 (CoF), the trained model demonstrated reliable predictive accuracy for WR, SWR, and FF. The respective corresponding RMSE and MAE values for WR and SWR were 0.000489 mm3/m and 0.000394 mm3/m, 0.000019 mm3/N·m and 0.000017 mm3/N·m, 0.2428 N and 0.1760 N, and 0.008357 and 0.006331, respectively. According to the SHAP interpretability analysis, sliding velocity mostly affected CoF and SWR, whereas load had the largest positive contribution to WR and FF. NSGA-II multi-objective optimization found pareto-optimal trade-offs that concurrently reduced wear and friction using the PINN as a surrogate model. With WR < 0.005 mm3/m and CoF = 0.33, the optimal parametric zones matched moderate loads (15–20 N) and mid-range velocities (2 m/s).
This study presents a pneumatically actuated wrinkled soft robotic pad (PAW-SRP) that enables active friction modulation for soft robotic fingers. Sharing the same pneumatic source as the finger, the pad changes its surface wrinkle morphology through pressure applied to internal chambers, increasing friction for stable grasping and reducing friction for compliant release. This design improves fingertip adaptability and grasp-release control without adding complex mechanisms, external stimuli, or intricate fabrication steps. Experimental results show that integrating the PAW-SRP with an existing soft finger increases the maximum graspable object weight to 355 g and enhances grasping performance. Compared with conventional passive or separately actuated surface designs, the proposed pad offers a simple, cost-effective, and versatile approach to tunable friction control, improving soft robotic finger functionality while maintaining compatibility with pneumatic actuation systems.
This study investigates head injuries sustained in car-bicycle collisions through multibody dynamics reconstruction coupled with finite element analysis (FEA). A collision dynamics model was employed to characterize the cyclist’s head kinematics under representative impact scenarios, accounting for variations in vehicle speed and impact angle. Anatomically accurate skull-brain finite element models were developed from medical imaging data and subsequently validated against experimental measurements. Utilizing crash reconstruction parameters, FEA simulations elucidated the propagation of stress waves through cranial tissues and quantified key biomechanical predictors of traumatic brain injury, including intracranial pressure, von Mises stress, shear stress, and cumulative strain. Results demonstrate that temporally-originating stress waves traverse both the skull and brain parenchyma, with localized stress concentrations exceeding critical injury thresholds correlating with varying severities of cerebral trauma. These findings establish a theoretical framework for the design of protective systems and advance the biome-chanical understanding of head trauma mechanisms in traffic accidents.
Aerosols pose a significant global challenge, prompting engineering advancements in filtration structures. This study investigates the impact of multilayer periodic lattice structures on the capture efficiency of aerosol, focusing on non-stochastic lattice designs. Two unit cell types were examined: a body-centered cubic (BCC) structure (BPL) and a hybrid face-centered cubic/BCC structure (CPL). Key design parameters, including cell type and filter layering, significantly influenced capture efficiency, airflow velocity, porosity, surface-to-volume ratio, and pressure drop. The CPL structure achieved pore sizes 80
