Aiming at the problems of high complexity in source domain data, inaccessibility of target domain data, and unknown fault patterns in real-world industrial scenarios for mechanical fault diagnosis, this paper proposes a Feature Decomposition-based Source-Free Universal Domain Adaptation (FD-SFUniDA) framework for mechanical equipment fault diagnosis. First, the CBAM attention module is incorporated to enhance the ResNet-50 convolutional network for extracting feature information from source domain data. During the target domain adaptation phase, singular value decomposition is applied to the weights of the pre-trained model’s classification layer, orthogonally decoupling the feature space into a source-known subspace and a target-private subspace. Then, based on the magnitude of feature projections, a dynamic decision boundary is constructed and combined with an entropy threshold mechanism to accurately distinguish between known and unknown class samples. Furthermore, intra-class feature consistency is strengthened through neighborhood-expanded contrastive learning, and semantic weight calibration is employed to reconstruct the feature space, thereby suppressing the negative transfer effect. Finally, extensive experiments under multiple operating conditions on rolling bearing and reciprocating mechanism datasets demonstrate that the proposed method excels in addressing source-free fault diagnosis problems for mechanical equipment and shows promising potential for practical engineering applications in fault classification tasks.

To address the insufficient prediction accuracy of traditional Arrhenius-type empirical correlations for hydrogen–oxygen combustion ignition delay time (IDT) over wide operating conditions, this paper proposes a data-driven prediction method based on Gaussian Process Regression (GPR). Based on 413 sets of shock tube and rapid compression machine experimental data, with temperature, pressure, and equivalence ratio as input features, an IDT prediction model was constructed. The GPR model adopts the Matérn 3/2 kernel function combined with the Automatic Relevance Determination (ARD) strategy for hyperparameter optimization. As a benchmark comparison, LASSO regression was employed for feature selection and parameter estimation of the Arrhenius empirical correlation. On a stratified sampling test set, the GPR model achieved a coefficient of determination R² of 0.8587 ($R_{ln}^2$ = 0.9271 on logarithmic scale), significantly outperforming the Arrhenius model’s R² of 0.5422 ($R_{ln}^2$ = 0.7027). In 50 random-split stability tests, GPR yielded a mean R² of 0.9180 with a standard deviation of only 0.0333, demonstrating excellent robustness. This study provides a valuable reference for ignition performance prediction of hydrogen–oxygen combustion systems.

Reliable formation tracking of fixed-wing unmanned aerial vehicles (UAVs) is particularly challenging when external acceleration disturbances, command saturation, actuator-induced acceleration tracking errors, sampled communication, and recurrent common-mode outages occur simultaneously. This paper develops a predictor-based robust event-triggered leader–follower framework for a three-dimensional flight-path point-mass outer-loop model. A projected adaptive distributed observer estimates the leader state and is shown to achieve practical fixed-time stability over effective communication-available intervals. A constant-velocity state predictor and a dimensionless sampled-data triggering rule reduce inter-UAV transmissions, where the outage model accounts for physical outages, candidate-grid and dwell-time waiting, and network resynchronization. Prediction-error terms of the same order as the nominal signed-power feedback are retained explicitly to derive sampled-mode dissipation conditions and uniform ultimate boundedness for bounded outages followed by sufficiently long effective communication-available intervals. An adaptive robust controller separately addresses the matched acceleration-channel uncertainty and the unmatched observer-correction input; the virtual inertial acceleration is subsequently mapped to commands for thrust, normal load factor, and flight-path bank angle. Numerical comparisons, ablation studies, sensitivity analyses, and integration-step convergence tests illustrate finite-horizon formation recovery and a communication–accuracy–control-effort trade-off without numerically certifying the analytical recovery-interval condition.

The cloud model provides a unified mathematical framework for representing both randomness and fuzziness simultaneously, offering an effective mathematical tool for modeling complex systems. This paper focuses on the emergent phenomenon of spontaneous synchronous flashing in fireflies and employs Gaussian cloud model theory to develop an initialization framework. Uncertainty is quantified using three core numerical characteristics of the cloud model—expectation, entropy, and hyperentropy—while a Gaussian potential function is adopted to characterize how interaction strength varies with spatial distance. A MATLAB-based simulation platform is developed to reproduce the emergent transition of fireflies from disordered flashing to synchronous behavior across different spatial domains. Our results demonstrate that group synchronization efficiency is strongly governed by the initial spatial configuration of individuals. Introducing driving nodes (special agents) can accelerate synchronization, but increasing their quantity yields diminishing marginal returns; the optimal strategy is a single special agent placed at the domain center. Collective emergence depends more strongly on the topological structure of inter-individual interactions than on the number of driving nodes. These findings confirm that interaction topology dominates synchronization performance over driving-node quantity, providing a theoretical basis for guiding unmanned swarm coordination with minimal control nodes.

An evidence-weighted multi-rate inertial fusion method is proposed for load displacement reconstruction in high-vibration industrial transportation scenarios, where inertial displacement reconstruction is readily affected by acceleration bias, attitude error, vibration disturbance, and false zero-velocity decisions. Multi-rate inertial data and high-bandwidth vibration/impact observations are used. The main inertial channel is used for motion-dynamics modeling, low-frequency inertial and attitude information is used to constrain the trajectory trend, and high-bandwidth vibration and impact responses are converted into motion-veto evidence so that unreliable stationary decisions under high-vibration conditions can be weakened. A Dempster–Shafer-style evidence fusion model is adopted to estimate motion, stationary, and unknown confidence, and the stationary confidence is propagated to ZUPT gating, covariance weighting, multi-branch displacement fusion, and keyframe factor-graph optimization. The experimental results show that stable load displacement reconstruction can be achieved by the proposed method, while interpretable diagnostic information, including displacement, velocity, stationary probability, ZUPT triggering, residuals, and drift risk, is also output. A diagnosable and auditable framework for inertial displacement reconstruction is therefore provided for high-vibration industrial scenarios where external references are difficult to deploy.

In response to the current limitation where conventional constant volume combustion apparatuses are generally confined to pressure ratings of 5–20 MPa, insufficient for the demands of ultra-high-pressure combustion fundamental research, this study designs and verifies a high-pressure-resistant constant volume combustion apparatus with a rated working pressure of 250 MPa. The strength design and safety factor calculation for the combustion chamber main body were conducted based on the Lame thick-walled cylinder elastic theory. A finite element numerical simulation method was systematically employed to perform static analysis, transient impact response analysis, and high-cycle fatigue-life assessment of the key components of the apparatus. The results indicate that under a 250 MPa design internal pressure load, the maximum circumferential stress at the inner wall of the combustion chamber main body is 328.0 MPa, with a safety factor greater than 1.5, complying with relevant safety codes for high-pressure vessels. Under transient loading simulating combustion impact, the maximum equivalent stress of all structural components is below the material yield strength, with a maximum elastic deformation of less than 0.06 mm, demonstrating excellent structural stiffness and impact resistance. Fatigue assessment with a design-life target of 1.0 × 10⁶ pressure cycles shows that the cumulative damage values for all components are significantly less than 1.0, meeting the reliability requirements for long-term cyclic service. This apparatus integrates functional modules such as high-pressure precision gas mixing, high-energy reliable ignition, high-speed transient parameter acquisition, and safe product collection, providing a stable, controllable, and safe experimental platform for in-depth research on the combustion mechanisms of gaseous fuels under ultra-high-pressure conditions.