Accurate bearing diagnosis requires informative vibration representations and reliable confidence estimates. Fixed evidence sets and independent fragment processing can overlook signal-dependent features and record-level relationships. This paper proposes a fragment-aware evidence large language model (FAE-LLM) that combines adaptive physical-evidence acquisition with trustworthy record-level fusion. A learned policy uses time-domain descriptors and waveform features to select spectral, envelope-spectrum, or autocorrelation evidence. Continuous projections map the selected evidence into a partially fine-tuned GPT-2 backbone. An order-aware relational transformer captures dependencies among adjacent fragments, and Dirichlet fusion integrates their evidence according to uncertainty and conflict. A validation-based rejection rule defers uncertain predictions. Across five runs, FAE-LLM achieves classification accuracies of 100 . 00 % on XJTU-SY and 96 . 35 % on GDUPT, with all GDUPT comparison methods receiving 4096-point records. Ablation studies show the benefit of longer record context, while fusion controls demonstrate improved probability calibration. The framework connects physical signal analysis with record-aware classification and selective decision support for bearing fault diagnosis.
Electronic Throttle Valve (ETV) actuators exhibit strong nonlinearities, friction layers, and tightly coupled electromechanical dynamics that severely challenge conventional controllers. This study introduces a rigorous Finite-Time stability Integral Sliding Mode Control (FTS–ISMC) architecture, augmented by an inner-loop current–shaping mechanism that leverages the inherent low-pass filtering dynamics of the DC motor to generate smooth and physically realizable armature currents. The mechanical subsystem is first stabilized through a finite-time integral sliding controller, which eliminates the reaching phase, guarantees robustness against matched disturbances, and enforces provable finite-time convergence of both the tracking error and sliding variable. The virtual control signal generated by the mechanical controller is subsequently embedded into the electrical subsystem via an analytically derived first-order current–tracking model, ensuring that the discontinuous control action is transformed into a bounded, low-chattering current compatible with actuator constraints. To further refine performance, Particle Swarm Optimization (PSO) is employed to tune the nonlinear gains, enabling multiple objectives such as minimizing the finite-time settling duration, reducing voltage demand, and achieving a balanced trade-off between speed and smoothness. Simulation results using the full nonlinear electromechanical ETV model confirm the effectiveness of the proposed framework: the FTS layer accelerates convergence; the ISMC layer enhances robustness and removes the reaching phase; the LPF-induced current shaping eliminates torque chattering; and PSO reduces settling time by 71% and decreases integral error indices by 52–78%. A sensitivity analysis of the electrical dynamics demonstrates that the nominal time constant yields the most stable and practically realizable behavior. These findings validate the feasibility of deploying the proposed FT–ISMC framework in real automotive throttle systems.
The current study examines the effects of various welding parameters on weld quality, as well as the mechanical and microstructural characteristics of austenitic stainless steel (ASS) 304, 316, and 201LN welds. Gas metal arc welding (GMAW), gas tungsten arc welding (GTAW), and Cold metal transfer (CMT) welding processes were used for investigation. Arc voltage, welding current, welding speed, wire feed rate, gas flow rate, and geometry of the weld bead were selected for investigation. Various optimization techniques like the RSM approach, grey-integrated Taguchi, and genetic algorithms were used to study the effect of process parameters. It was observed that the medium value of the arc voltage (18.5 V) significantly affected the micro-hardness, with the maximum hardness of 268.86 VH in all welded joints with a wire feed rate of 116 mm/s and WD of 0.8 mm, while the wire feed rate (137 mm/s) had the biggest impact on the tensile strength (662.8 MPa) of the joints with an arc voltage of 19 V and WD of 1 mm. The material’s tensile and bending strengths are influenced by current, speed, and root gap. The HAZ of every joint exhibited evidence of grain coarsening. The best quality weld was found at a low level of heat input (1.25 kJ/mm) with maximum UTS and toughness. It has also been observed that grain coarsening occurs as heat input rises. The effect of gas flow rate on weld quality was also assessed.
Minimum noise amplitude deconvolution (MNAD) has shown considerable potential for enhancing impulsive signatures in rolling bearing fault diagnosis. Nevertheless, its effectiveness is highly sensitive to the selection of prior parameters, particularly the fault period. In engineering practice, the parameter is commonly obtained either from manual calculations based on shaft speed and bearing geometry or from empirical experience, which restricts the adaptability and automation of the method. To overcome this limitation, a fault-period-adaptive minimum noise amplitude deconvolution (FPA-MNAD) approach is proposed in this study. Firstly, a multiscale envelope spectrum dominant peak detection (MESDPD) scheme is proposed. Based on the MESDPD, the fault characteristic frequency is determined and further transformed into the corresponding fault period, avoiding manual derivation and subjective threshold selection. The estimated fault period is then integrated into the MNAD procedure, allowing adaptive updating and optimization of the filter period parameter. In this way, weak repetitive transients and their periodic behavior can be extracted more effectively in heavy noise environments. Experimental investigations demonstrate that the proposed method provides impressive fault feature enhancement and improved diagnostic robustness, while reducing reliance on prior information and operator intervention. These results indicate that the method has strong potential for practical engineering applications.
Thermal management remains a critical challenge in contemporary energy and hydrogen-based energy systems, especially when considering fuel cells, hydrogen compressors, and compact heat exchangers in which high temperatures decrease efficiency and reliability. Motivated by the combined effects of thermophysical properties, non-Newtonian rheology, particulate contamination, porous resistance, and magnetic control, this study investigates MHD flow of an engine-oil-based Cu-Fe 3 O 4 hybrid nanofluid through an inclined porous channel. Cu enhances thermal transport, while Fe 3 O 4 enables magnetic control of the flow. The study considers an unsteady two-phase dusty Casson fluid model between inclined parallel plates under the influence of a relative magnetic field, taken to be fixed either with respect to the fluid (MFFRF) or with respect to the plate (MFFRP). Additionally, porous medium resistance, thermal radiation, and internal heat generation are incorporated to represent practical thermal systems. The governing momentum and energy equations are solved analytically using the Poincaré-Lighthill perturbation technique (PLPT). Thermal and flow characteristics are evaluated using engineering performance indicators, namely the skin-friction coefficient and the Nusselt number. From the results, it is observed that the heat transfer rate of the hybrid nanofluids is higher than that of the mono-nanofluids and conventional engine oil. When the volume fraction of the nanoparticles varies from 0.01 to 0.04, the value of the Nusselt number varies from 2.10 to 2.80 for the conventional oil, from 2.18 to 3.11 for the Cu nanofluid, and from 2.64 to 3.94 for the Cu-Fe 3 O 4 hybrid nanofluid, showing a growth of 25.71%–40.71%. Lorentz braking and Darcy resistance suppress momentum transport, whereas buoyancy and channel inclination enhance convective heat transfer. The results can prove useful for designing magnetically controlled cooling passages and heat-exchanging devices for use in hydrogen fuel cell cooling, hydrogen compressors, and other hydrogen energy systems.
