Accurate rolling-bearing remaining useful life (RUL) prediction is important for reliable equipment operation. This study proposes an RUL prediction framework integrating cumulative damage representation, adaptive degradation identification, and intelligent optimization. Time-domain features are cumulatively summed along the lifetime axis, and a multi-scale temporal convolutional autoencoder (MS-TCN-AE) is used to construct a health indicator (HI). The adaptive threshold maximum rate of change (ATMROC) algorithm then estimates the initial degradation time (IDT) and extracts the degradation stage. Finally, BiGRU-MHSA predicts RUL, while Student Psychology-Based Optimization (SPBO) optimizes key hyperparameters using validation error. Experiments on the PHM2012 and XJTU-SY datasets verify the effectiveness of the method. Independent retraining on the NASA IMS dataset further indicates adaptability to different data sources and fault types.
Unmanned underwater vehicles (UUVs) are indispensable to modern marine operations, performing critical tasks such as anti-submarine reconnaissance, underwater resource exploration, and multi-platform collaborative search-and-rescue. As underwater communication system serves as the fundamental basis for mission success for UUV systems, generally, their reliability and safety of it should be estimated during the mission planning phase. The reliability and safety estimation of the communication link is a precondition for that of the system, and therefore this paper establishes a comprehensive evaluation model integrating link reliability and safety: it adopts the BELLHOP model to rectify limitations of traditional link analysis methods, enhancing its adaptability to complex marine environments, and combines the Monte Carlo method for link task reliability assessment. An illustrative shallow-sea case verifies that the proposed model outperforms traditional path loss-based methods in prediction accuracy, thus confirming its validity for UUV communication link evaluation. To demonstrate the engineering value of the proposed model, an underwater search-and-rescue backhaul communication application is adopted.
The ship lubrication system is susceptible to multiple influencing factors—including temperature, pressure, oil quantity, and oil quality—under complex operating conditions, leading to nonlinear fault evolution patterns that make traditional monitoring methods inadequate for timely and accurate early warning. To address these challenges, this study proposes an intelligent maintenance model integrating an improved Dynamic Bayesian Network (DBN) with a Convolutional Neural Network-based Bidirectional Gate Recurrent Unit (CNN-BiGRU). The approach enhances system evolution modeling capabilities through optimized dynamic state transition mechanisms, utilizes CNN-BiGRU to extract multivariate time-series features, and combines probabilistic inference for abnormality detection and risk assessment. Experimental results on the Open Ship Machinery Data Set (SMD) and Oil Analysis Data Set (OAD) demonstrate that the proposed model achieves an F1-score of 95.52% and an AUC of 0.971, outperforming multiple competing models. Furthermore, the model exhibits superior early warning performance and stability under various abnormal operating conditions and load fluctuations. These findings confirm that the proposed methodology significantly improves fault monitoring accuracy and intelligent maintenance efficiency for ship lubrication systems, providing robust technical support for reliable equipment operation under complex environments
This paper presents a comprehensive mathematical and statistical analysis of the generalized Gompertz and reversed generalized Gompertz distributions, introduced by [W. J. Willemse and R. Kaas, Rational reconstruction of frailty-based mortality models by a generalization of Gompertz’ law of mortality, Insur. Math. Econ. 40 (2007) 468–484]. A wide range of mathematical properties is rigorously derived, including the shapes of the density functions, moment generating and characteristic functions, ordinary and conditional moments, quantiles, mean deviations, mean residual life, expected inactive time, inequality curves (Lorenz, Bonferroni, Zenga), order statistics, extreme value limits, and Kullback–Leibler divergence measures (Shannon, Rényi, Tsallis, Kullback–Leibler) and the distributions of the sums and differences of generalized Gompertz and reversed generalized Gompertz random variables. The maximum likelihood method is employed for parameter estimation, and the asymptotic properties of the estimators are established. A detailed simulation study demonstrates the finite-sample performance of the estimators, confirming their consistency, low bias, and adequate coverage probabilities across various parameter configurations and sample sizes. Finally, the applicability of both distributions is illustrated using a real data set on lifetimes of glass capacitors, where both models provide satisfactory fits and outperform a range of standard competitors. The results indicate that the distributions are flexible and theoretically well-founded models suitable for applications in reliability.
