Photovoltaic (PV) systems, particularly in hot climates. This study experimentally investigates nanofluid-based cooling as an effective strategy to enhance PV performance under real outdoor conditions. Two environmentally friendly nanofluids—ferromagnetite (Fe 3 O 4 ) and calcium carbonate (CaCO 3 )—were examined at concentrations ranging from 0.5 to 1.3 wt.% and flow rates between 1000 and 7000 ml/min, with water cooling and a non-cooled PV panel used as reference cases. The results show that Fe 3 O 4 at 1.3 wt.% and the highest flow rate achieved the best overall performance, yielding a temperature reduction of 36.3%, a power output increase of 27.6%, and an electrical efficiency enhancement of 27.8%. CaCO 3 also demonstrated notable performance, achieving a 34.77% temperature reduction and approximately a 20.52% increase in power output, and a 27.84% increase in efficiency, while water cooling provided only moderate improvements. Increasing nanoparticle concentration and flow rate under turbulent conditions (1.3 wt.% and 7000 ml/min) significantly enhanced the performance of Fe 3 O 4 nanofluid, which exhibited a peak Nusselt number of 110.23, substantially higher than those of CaCO 3 (55.27) and water (15.32). In addition, the thermal performance factor increased, indicating improved overall heat transfer characteristics. The selection of Fe 3 O 4 and CaCO 3 is motivated by their chemical stability, non-toxicity, low cost, and availability, supporting both environmental and economic sustainability. Overall, the results confirm that nanofluid-based cooling significantly outperforms conventional cooling methods, offering a scalable and climate-resilient solution for enhancing PV system efficiency.
To address contact ablation, mechanical wear and maintenance challenges in conventional mechanical driver controllers used in freight diesel locomotives, this study develops a digital-retrofitting scheme that replaces the mechanical control structure while retaining the original drive-system architecture. A host–slave dual-decoupled architecture based on an isolated RS-485 bus and the Modbus-RTU protocol is implemented. The STM32F429-based host unit integrates a human-machine interface (HMI) and software interlocking, while the STM32F103-based slave unit drives a ULN2003ADR array and a 19-channel industrial relay array to reproduce the coding logic of conventional cam contacts at the ADLC-4 interface. Bench-scale validation on an integrated test platform demonstrated a driver-circuit voltage margin of 0.4 V and a current margin of 484.9 mA. The system responded correctly to directional and 33-level speed-regulation commands, and the software interlock rejected the tested invalid input sequences. These results provide bench-scale evidence supporting the feasibility of digital retrofitting for legacy freight-locomotive driver controllers. Field-test performance, long-term reliability, life-cycle cost, and environmental robustness require further validation.
Traditional fault diagnosis methods suffer from performance degradation when applied under varying operating conditions. This presents a critical challenge to the analysis of industrial systems, where distribution shifts in sensor data severely degrade diagnostic reliability. Domain adaptation approaches have shown promise in cross-condition scenarios, but current methods face two fundamental limitations: (1) insufficient feature hierarchy for capturing multiscale fault signatures and (2) incomplete domain alignment, neglecting conditional distribution discrepancies. This paper proposes a Time-Frequency Hybrid Multilevel Feature Alignment Domain Adaptation Network (TF-MFA-DA) that integrates three coordinated design components: (a) a time-frequency hybrid feature extractor, in which the time-domain branch uses multiscale convolution and Transformer encoding to capture local impulsive patterns and global temporal dependencies, while the frequency-domain branch extracts spectral-statistical information; (b) a dual-alignment mechanism enforcing both marginal and class-conditional distribution matching through global and conditional domain discriminators; and (c) a weighted maximum mean discrepancy (WMMD) based adaptive feature weighting module for enhanced cross-domain transferability. Extensive experiments on two publicly available datasets demonstrate that the proposed framework achieves competitive diagnostic performance and improved stability under diverse operating conditions.
The dynamic stability of guideway protection devices directly affects the accuracy and reliability of high-speed machine tools. To address instability caused by inertial impact and structural vibration, this study investigates three aspects: transmission mechanism selection and link parameter optimization, nonlinear analysis of joint clearance, and modal and transient response analysis of cover plates. The synchronous parallel mechanism reduces the maximum stress by 83.08% compared with the synchronous mechanism. The optimal link width-to-thickness ratio is 0.83 (10 mm × 12 mm). As the joint clearance increases from 0.1 to 0.8 mm, the peak impact force rises from 60 to 318 N, and the impact pattern transitions from dense to sparse contact; clearance should be controlled within 0.2 mm (upper limit 0.4 mm). Modal and transient analyses show the apex angle has no significant effect on natural frequency but notably affects transient deformation. The 175° cover plate exhibits the lowest deformation and stress under a 155 N impact, with reductions of 27.96% and 31.82% relative to the 170° scheme, balancing dynamic stability and chip evacuation. Thus, adopting the synchronous parallel mechanism, controlling clearance within 0.2 mm (upper limit 0.4 mm), and using the 175° cover plate effectively enhances dynamic stability.
To address the small-sample fault diagnosis challenge of gas turbine rotors under complex operating conditions, a Conditional Residual Generative Adversarial Network incorporating a Self-Attention Mechanism (CRGAN-SAM) is proposed. Residual modules are embedded in the generator and discriminator, conditional label vectors guide fault-category expansion, and SAM strengthens global time-frequency feature capture. Ablation experiments showed that spectral normalization combined with Leaky ReLU at a learning rate of 0.20 was optimal, yielding a Fréchet Inception Distance of 15.22 and a Maximum Mean Discrepancy of 0.083. On the simulation test set, CRGAN-SAM achieved 99.92% accuracy for the normal state and more than 99.6% for support looseness, rotor rubbing, and imbalance. In an independent field dataset from two operating gas turbine units, the model achieved 95.83% accuracy and a macro-F1 of 95.83% without parameter updating, outperforming TMCGAN and MPAGAN. The results indicate that CRGAN-SAM maintains high sample-generation fidelity and diagnostic robustness under small-sample and field conditions.
