7 results about "Structural decomposition" patented technology

The invention relates to the technical field of data compression, in particular to a ZSTD compression algorithm optimization method based on joint coding and ternary decomposition, and the method comprises the following steps: S1, building a joint frequency table based on the joint distribution of the literal length and the matching length of a previous section, setting a frequency deviation criterion, and dynamically adjusting a low-frequency combination proportion; s2, decomposing the offset value structure into a leading segment, a trailing segment and a middle segment, performing bit width-entropy ratio analysis on the middle segment, and selecting a symbol index or bit stream coding; and S3, entropy discriminant coding is performed on the joint symbol and the offset symbol respectively, and a structure position is mapped synchronously to generate a structured compressed output stream. According to the method, collaborative optimization of frequency self-adaption, redundancy minimization and structured mapping is realized through joint symbol dynamic modeling and offset ternary structure entropy discriminant coding, so that the ZSTD compression efficiency and the coding stability are remarkably improved.

The application discloses a kind of object grid generation method and system based on three-dimensional model, the method includes: obtaining the three-dimensional model of object;The segmentation boundary is determined by carrying out segmentation operation to three-dimensional model;Based on the segmentation boundary, cut three-dimensional model into multiple segmentation regions;Boundary zone is generated between adjacent segmentation regions;The grid region of the boundary zone is executed to realize boundary reconstruction by remeshing operation, and the reconstructed segmentation boundary is obtained;Based on the reconstructed segmentation boundary, generate the update grid of object;And between detachable component and fixed component separated by segmentation boundary, build multiple connection structures.The application realizes structure decomposition and boundary reconstruction optimization based on the three-dimensional model of object, improves the structure compatibility, assembly stability and connection transition smoothness in additive manufacturing process, improves the rationality and manufacturing adaptation ability of reconstructed morphology.

This invention relates to the field of hydrological model parameter calibration technology, specifically to an automatic parameter calibration method and system for a surface water-groundwater coupled model. The method includes: acquiring multi-dimensional measured and model simulation output sequences of the surface water and groundwater subsystems; fusing and constructing coupled hydrological data entities containing time-varying and spatial distribution characteristics; separating unstable response features based on spatiotemporal deviations after applying perturbation; filtering out these features to form a de-perturbed steady-state data entity and matching it with a multi-mode parameter response template library; generating a hierarchical model parameter feature cluster set through adaptive structural decomposition; constructing a dynamic constraint network for surface water and groundwater parameters; obtaining a preliminary adjustment scheme through synchronous optimization and cross-validation; locating key transmission and boundary interaction parameter nodes; optimizing the scheme path; and generating the final parameter calibration instruction sequence. This method eliminates non-steady-state data interference, adapts to the parameter correlation characteristics of the coupled system, and achieves adaptive and precise execution of parameter calibration.

The invention discloses a method for automatically generating a beam and a plate based on column structure decomposition and axis net connection. According to the method, geometric information of columns is extracted from a two-dimensional structure planar graph, through the processes of axis net decomposition, adjacent coupling beam pairing, degree sealing and surface domain tracking plate forming, a beam net frame is automatically generated according to the positions of the columns, a floor slab surface is finally generated, and therefore the complete topological relation among the columns, the beams and the slabs is established. According to the method, generation of wrong beam segments or wrong connections can be effectively avoided through a pairing strategy of pseudo interval filtering and outer contour priority. Therefore, the stable automatic modeling process of automatically generating the beam from the column and automatically generating the plate from the beam is realized. The method is used in a two-dimensional building structure plane graph, the arrangement of beams is automatically generated according to the positions and shapes of columns, and the floor outline is formed through closing. According to the method, manual participation can be reduced, manual omissions and errors are avoided, the modeling efficiency is greatly improved, and an accurate data basis is provided for engineering quantity calculation and cost analysis.

The invention discloses a competing product comparison method based on a knowledge graph, and the method comprises the following steps: obtaining competing product comparison multi-source structural data, and constructing a competing product comparison knowledge graph and a competing product entity pair input structure; in combination with a CompGCN model, constructing a competing product alignment graph reasoning framework and a relation combination operator, and generating a propagation record; performing structural decomposition and recombination on the relation combination operator, constructing dimension convolution configuration, and generating a dimension competing product representation result; generating a difference propagation path set based on the dimensionalized competitive product representation result and the propagation record; mapping the difference propagation path set into a comparison feedback signal, and modulating a competing product alignment graph reasoning framework; in the modulated competing product alignment graph reasoning framework, feedback alignment convolution propagation is executed, and a competing product comparison representation result is generated; and generating a competitive product comparison conclusion based on the competitive product comparison representation result. According to the invention, traceable modeling and stable comparative analysis of the competitive product difference forming process are realized.

The invention relates to the technical field of artificial intelligence, and provides a model training method and a related device, and the method comprises the steps: carrying out the structure decomposition of an initial model based on a model structure of a reference model, and determining a parameter updating structure network in the initial model and a parameter inheriting structure network which is the same as the model structure of the reference model; inheriting model parameters of the reference model by using the parameter inheriting structure network, performing knowledge distillation on the initial model based on the reference model to obtain updated parameters of the parameter updating structure network, and performing parameter updating on the parameter updating structure network in the knowledge distillation process; generating target model parameters according to the model parameters of the reference model and the updated parameters, and updating the model parameters of the initial model based on the target model parameters to obtain a target model. Through parameter inheritance, the parameter integrity of the reference model is ensured, and rapid transmission of knowledge is realized in combination with knowledge distillation, so that the model training speed and the model performance are improved.

The invention discloses a micro-grid load prediction method based on deep learning, and the method comprises the following steps: S1, collecting micro-grid load data, and generating a load time sequence; s2, executing load structure decomposition to generate load subsequences; s3, executing load operation scene identification, and generating a load operation scene label and a load structure subsequence; s4, inputting the load structure sub-sequence into an improved SCINet model, introducing a load structure perception guide mechanism, and outputting a load prediction sequence; s5, executing consistency constraint processing, and generating a consistency load prediction sequence; s6, generating prediction deviation data; and S7, executing adaptive updating, and outputting an updated load prediction sequence. According to the method, high-precision prediction and continuous adaptive updating of the micro-grid load time sequence under the conditions of multiple operation scenes and multiple time scales are realized, and the stability and reliability of a load prediction result are effectively improved.