System identification combined with free-running model tests or full-scale trials is one of the effective methods to determine hydrodynamic coefficients in mathematical models ship maneuvering motion. By analyzing available data, including rudder angle, surge speed, sway yaw rate, and so forth, a method based on support vector machines (SVM) estimate proposed for conventional surface ships. The are contained expansion inner product linear kernel function. Predictions motion conducted by...

Combined with the free-running model tests of KVLCC ship, system identification (SI) based on support vector machines (SVM) is proposed for prediction ship maneuvering motion. The hydrodynamic derivatives in an Abkowitz are determined by Lagrangian factors and vectors SVM regression model. To obtain optimized structural SVM, particle swarm optimization (PSO) incorporated into SVM. diminish drift after regression, a difference method adopted to reconstruct training samples before...

A fusion algorithm is proposed for the restoration and enhancement of underwater images. Color balance, contrast optimization histogram stretching are carried out. To alleviate effect color shift in an image, scalar values R, G, B channels renewed so that distributions three similar. Instead refining transmittance dark channel prior based restoration, optimized employed by which optimal determined. further improve brightness images, a on red given. verify effectiveness algorithm,...

The YOLOv4 neural network is employed for underwater target recognition. To improve the accuracy and speed of recognition, structure modified by replacing upsampling module with a deconvolution incorporating depthwise separable convolution into network. Moreover, training set used in YOLO preprocessed using mosaic augmentation, which gray world algorithm to derive two images when performing augmentation. recognition results comparison other detectors demonstrate effectiveness proposed method...

Due to refraction, absorption, and scattering of light by suspended particles in water, underwater images are characterized low contrast, blurred details, color distortion. In this paper, a fusion algorithm restore enhance is proposed. It consists restoration module, an end-to-end defogging module brightness equalization module. the balance based on CIE Lab model proposed alleviate effect deviation images. one end input image other output image. A CNN network connect these two ends improve...

A model of a submerged angular cavitation nozzle is established, which consists contraction part, parallel middle and expansion part. Based on the CFD technique, numerical simulation flow field carried out, in multiphase mixture model, renormalization group (RNG) k-ε turbulence are applied. Considering influence density cavitation, effects inlet outlet part velocity vapor volume fraction studied. The results show that essential jet. When diameter d fixed, designed with angle α = 13.5°,...

This paper explores the application of multidisciplinary design optimization to blades in horizontal-axis wind turbines. The aerodynamics and structural performance are considered framework. In aerodynamic discipline, class function/shape function transformation-based parameterized modeling is used express airfoil. Wilson method employed obtain shape blade. Computational fluid dynamics numerical simulation performed analyze materials ply lay-up studied. Finite element method-based modal...

To improve the feasibility of system identification in prediction ship manoeuvrability, several measures are presented to deal with parameter identifiability parametric modeling manoeuvring motion based on identification. Drift nonlinear hydrodynamic coefficients is explained from point view regression analysis. diminish multicollinearity a complicated model, difference method and additional signal employed reconstruct samples. Moreover, structure model simplified correlation Manoeuvring...

To reduce the ship wave-making resistance, lines of bulbous bow a hull are optimized by an automatic optimization platform at design stage. Parametric modeling was applied to using non-uniform rational basis spline (NURBS). The Rankine-source panel method used calculate resistance. A hybrid strategy achieve goal. Ro-Ro taken as example illustrate adopted, with objective minimize results show that resistance obviously reduces and wave-shape near becomes gentle after optimized, which...

Based on deep learning, an underwater image instance segmentation method is proposed. Firstly, in view of the scarcity related data sets, size set expanded by measures including rotation and flipping, generation a generative adversarial network (GAN). Next, finally constructed manual labeling. Then, order to solve problems color shift, blur poor contrast optical images caused complex environment attenuation scattering light, enhancement algorithm used first preprocess set, several algorithms...

Navigating through complex flow fields, underwater vehicles often face insufficient thrust to traverse particularly strong current areas, necessitating consideration of the physical feasibility paths during route planning. By constructing a field database Computational Fluid Dynamics (CFD) simulations operational environment, we were able analyze local uncertainties within field. Our investigation into path planning using these data has led proposal hierarchical strategy that integrates...

This paper presents a robust cascaded control strategy to underwater robot thrust. The dynamics of surge motion, propeller axial flow, shaft, and electrically driven circuit in the motor constitute system with respect Instead usual parameter perturbation, generalized modeling errors are considered plant, which may be parametric errors, ignored high-order modes, or some unmodeled thrust system. External disturbances also taken into account, random noises from mechanical electrical equipment,...

We study the problem of learning a single occurrence regular expression with interleaving (SOIRE) from set text strings possibly noise. SOIRE fully supports and covers large portion expressions used in practice. Learning SOIREs is challenging because it requires heavy computation usually contain noise Most previous studies only learn restricted are not robust on noisy data. To tackle these issues, we propose noise-tolerant differentiable approach SOIREDL for SOIRE. design neural network to...