4results about How to "High numerical precision" patented technology

The application discloses a kind of double box girder flow field dynamics modal and flow stability analysis method and system, belong to bridge engineering field.This method first establishes double box girder flow field model by computational fluid dynamics, solves the momentum equation and continuity equation of two-dimensional incompressible flow;Dynamic modal decomposition is carried out again, and eigenvalue, characteristic mode and other information are obtained;Finally, based on linear stability theory, the flow field component is expressed as the sum of steady-state solution and disturbance term, and the global linear flow stability equation is obtained by substituting into the equation, and the flow stability is analyzed by discrete.This method can accurately extract the modal characteristics of each order, providing a scientific basis for wind-induced vibration analysis of double box girder, and has important application value.The method of the application occupies less computing resources, has high numerical accuracy, and can be further expanded to higher precision calculation model according to computing resources, and is suitable for application in actual bridge engineering.

The present application belongs to the field of ship hydrodynamics and motion prediction, and in particular to a device and method for predicting the hydrodynamics and motion of a ship coupled with ice, which first derives a Green function that satisfies the boundary conditions of the ice layer and the free surface, as well as all other conditions except the surface condition of the ship, and through the Green function, the differential equation of the velocity potential is converted into a boundary integral equation on the surface of the ship only, which introduces an extended surface at the waterline of the ship body to eliminate the influence of irregular wave frequencies; an asymptotic solution formula of the Green function is analytically derived, thereby establishing an approximate and effective solution program for the coupled interaction of the ship / wave / ice layer, and numerically calculating the results of the added mass, damping coefficient and wave exciting force of the ship body; through the method, the interaction mechanism of the water wave / ship / ice layer under different parameter characteristics can be deeply understood, and theoretical support is provided for the hydrodynamic problems, motion prediction and rational design of polar ships.

This application discloses a machine learning-based method for constructing numerical atlases of marine propellers, relating to the field of marine technology. This method uses a viscous flow numerical calculation model to calculate the performance parameters of a propeller with design parameters for each set of samples at various sample infeed velocities. Then, using the design parameters and infeed velocities of each set of samples as input and the corresponding performance parameters as output, a performance surrogate prediction model is trained using multiple sets of sample design parameters based on a machine learning model. This surrogate prediction model can then be used to obtain the performance parameters corresponding to different propeller schemes within the atlas parameter range, thus constructing a numerical atlas of marine propellers. This method combines the high efficiency and high accuracy of viscous flow numerical methods with the superior nonlinear fitting capability of machine learning models. It does not rely on extensive physical model experiments, has a simple construction method, and produces propeller atlases with high numerical accuracy.

This invention discloses a method for processing matrix multiplication weight data in model inference, adapted to large-scale matrix multiplication in BFP format for large models in the Transformer architecture. Addressing the pain points of existing architectures regarding BFP exponent and mantissa splitting, bandwidth, and resource consumption, this method employs (64,64) weight blocks, dual address generator 1:4 scheduling, waterline flow control, and double buffering techniques to achieve full-bandwidth read / write between DDR, BRAM, and computational units. Through multi-core parallelism and multi-round multiplication and accumulation, performance is improved by more than 3 times compared to FP16, reducing memory usage and design complexity while ensuring inference accuracy, providing an efficient hardware acceleration solution for LLM inference in resource-constrained environments.