In this article, a novel prediction index is constructed, hybrid filtering proposed, and remaining useful life (RUL) framework developed. the proposed framework, different models are built for operation states of rolling bearings. normal state, linear model built, Kalman filter (KF) implemented to determine failure start time (FST). degradation dimensionless CRRMS based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) wavelet threshold. Then, double exponential...

The failure threshold of a product is not always constant when affected by random shocks. Thus, this paper establishes reliability model that combines dynamic and self-healing characteristics, which are subjected to the competing process. Soft caused degradation exceeding soft threshold, hard occurs because Degradation consists natural increment thresholds rate will also be Moreover, characteristics in process, together with increment, affect whole When shock arrives, change simultaneously....

In this paper, a fault diagnosis method is proposed based on multi-sensor fusion information for single and composite of train braking systems. Firstly, the mass model brake established operating environment. Then, pre-allocation linear-weighted summation criterion are to fuse monitoring data. Finally, improved expectation maximization, modes parameters identified, faults diagnosed in real time. The simulation results show that systems can be effectively accurately identification results....

Given multiple degradation failure and shock processes within the complex system during operation, a reliability model that combines degradation-shock competing process dynamic threshold is modeled. The Wiener with random effects considered model, which includes to account for heterogeneity among units. Then, extreme used depict shock. copula function carried out illustrate correlation between processes, constructed further. To verify application of this numerical case...

The samples representing abnormal situation is usually very few in the dataset, which makes it difficult to learn features of by machine-learning-based methods. To improve accuracy anomaly detection, number should be expanded ensure balance dataset. In this paper, a discrete synthetic minority oversampling technique (D-SMOTE) proposed generate new samples. A closed area constructed using three nearest are then uniformly interpolated area. By means, problem imbalance for original dataset...

Traffic accidents occurred frequently on roads, which bring huge losses to society. The purpose of this article is extract the important influence factors traffic accidents, reduce probability road and support basis for decision-making unmanned vehicles. For all this, an improved artificial neural network method proposed predict analyze severity vehicle accidents. impact different factors, such as road, weather, surface, time, etc., accident huge. Hence, are quantified redundancy between...

Abstract The degradation process analysis is the basis of system reliability assessment. Too few samples will affect accuracy analysis. So, it necessary to expand data and construct prediction model. This paper proposes an SVAE‐GRU‐based generation model handle issues. New are generated by combining Stacked Variational Autoencoder (SVAE) Gated Recurrent Units (GRU) in each time step, which can improve learning effectiveness features. GRU Multi‐layer perceptron calculate mean variance...

Considering the competitive failure that exists in operation of industrial systems, including degradation and sudden failure, this paper presents a reliability assessment method based on three-parameter Weibull distribution Wiener process. Nonlinear functions are proposed to establish relationship between different processes, model is derived. The Metropolis-Hastings (MH) sampling algorithm Monte Carlo Markov Chain (MCMC) employed estimate parameters study. results obtained by real samples....

Abstract For nonlinear stochastic systems, controlling low‐order moment indexes, such as mean and variance, may not effectively achieve actual control requirements since they do reflect the complete statistical characteristics of random variables. Instead, probability density function (PDF), which provides characterization properties, can be used to each order more efficiently. Existing PDF shape methods are mainly suited for systems with polynomial forms, leads limitations in their...

Abstract Robot visual place recognition is a research hotspot, and robot systems based on 3D point clouds are even more focused. In recent years, scan-context descriptor imagery has become standard method for cloud positioning. Based its system characteristics, we integrated it with the Shuffle Attention (SA) mechanism module to improve performance. And tested public database NCLT dataset (North Campus Long-Term Vision dataset), our good results.

<title>Abstract</title> Visual place recognition using 3D lidar in robotics is a popular research topic. Feature representation descriptors, such as scan-context which are 2D descriptors obtained from point clouds, one of the directions with great in-depth value. This paper proposes new method to improve robot visual based on descriptors. Our approach focuses combination loss functions and optimizers accuracy robustness system. Specifically, we introduce custom function that encourages...

In this paper, a method for calculating vehicle collision probability is proposed. The running state and the relationship between velocity displacement are determined by analyzing speed sequence. Using Monte Carlo simulation method, frequency histogram of time that arrived at point obtained. order to get distribution function, corresponding fitted Gaussian fitting method. At last, two vehicles intersection calculated formula. experiment results show can calculate exactly, which provide...

It is essential to assess the autonomous vehicle operation safety during driving, which can avoid or reduce collision risk by evaluating drive safety. In this paper, a assessment algorithm proposed, that quantifies auto-drive Time Collision (TTC) frequency. Firstly, Long Short Memory network (LSTM) used predict surrounding trajectory; Moreover, point between and determined, frequency distribution result of TTC calculated Monte Carlo simulation method; Finally, running speed & hazard...

The frequent occurrence of traffic accidents has brought huge losses to society. It is necessary predict the accident severity in advance. Since different, distribution their categories generally imbalanced, which will increase running risk vehicle. To solve this problem effectively, an improved sampling method proposed paper. Firstly, a map encoding used samples into discrete data. Secondly, Discrete Synthetic Minority Oversampling Technique (D-SMOTE) generate that are close values....

Since the cumbersome collection process and high cost, collected degradation of product is basically small samples, which will affect accuracy reliability evaluation. It necessary to expand improve later assessment. Therefore, a generation prediction method proposed combining time series generator adversarial network (TimeGAN) stochastic process. Firstly, input expanded by sliding window training accuracy; Then, construction in TimeGAN linked with make data more realistic. Finally, results...

Based on the reliability analysis of correlated competitive risk systems with degradation processes and stochastic shocks, a dependent failure model considering time-varying soft threshold is proposed The process includes caused by continuous hard random impact. In complex equipment, when system degrades faster due to performance degradation, will decrease performance. this study, proposed. Through numerical micromotor, overestimated certain extent comparing models without change threshold.