17results about How to "Improve structural rationality" patented technology

The invention discloses a multi-modal semantic topology and constraint reasoning CAD sequence generation method based on a time scale, and belongs to the technical field of crossing of computer aided design and artificial intelligence. According to the method, an end-to-end automatic generation framework from multi-modal input to an editable parameterized CAD model is constructed, and semantic consistency, structural rationality and logic correctness of a generated sequence are remarkably improved; according to the method, multi-source data such as images, sketches and point clouds can be deeply fused, a design intention is accurately captured through feature extraction of time scale perception and dynamic topological graph evolution, and meanwhile, the problems of error accumulation and logic drift in a traditional autoregression model are effectively solved through a sequence synthesis mechanism guided by grammar and geometric constraints; and in combination with a geometric feedback driving optimization engine, efficient and steady generation of the parameterized model is realized, so that systematic breakthrough in the aspects of generation quality, automation degree and cross-scene applicability is realized, and a key technical support is provided for intelligent manufacturing and intelligent design.

This invention relates to the field of image generation technology, and more particularly to a method for generating bird's-eye view images of reservoir landscapes based on a Stable Diffusion large model. The method includes the following steps: preprocessing acquired reservoir landscape images and semantically annotating them based on water conservancy terminology to construct an image-text pairing dataset; training a pre-trained Stable Diffusion model, incorporating LoRA adaptation parameters into its UNet attention layer, using the image-text pairing dataset to generate a target feature image; and updating the LoRA adaptation parameters based on the difference between the target feature image and the reservoir landscape image to obtain a LoRA model for the reservoir landscape bird's-eye view image. This invention achieves structurally controllable and semantically consistent reservoir landscape generation by fusing structural reference constraints and semantically weighted alignment and introducing LoRA modulation.

This utility model relates to the technical field of anti-collision pipe processing equipment, specifically a rotary fixture for circumferential welding of anti-collision pipes. It includes a fixture base frame, with an L-shaped support frame symmetrically fixed to the top of the base frame near its center. A through hole for the support plate is opened at the center of the side wall of the L-shaped support frame. Clamping assemblies are symmetrically installed above the fixture base frame, each clamping assembly including two positioning ring plates. The clamping and fixing of the anti-collision pipe to be welded is achieved through the mutual cooperation between the structures belonging to the clamping assemblies on both sides of the fixture base frame. The two positioning ring plates are respectively installed inside the through hole of the support plate and the sliding ring plate. This allows the anti-collision pipe to rotate as a whole with the positioning ring plates, the expansion connecting rod, and the extrusion clamping plate after the positioning ring plates clamp and fix the anti-collision pipe close together, replacing the rotation process of the welding torch during circumferential welding. This results in higher operating efficiency and increases the working efficiency of circumferential welding of anti-collision pipes.

This invention discloses a data and knowledge fusion-driven method for generating wafer image defect detection models. First, a process-defect-module adaptation knowledge graph is constructed, and defect feature condition vectors and module priority vectors are generated based on directed knowledge reasoning along the defect-process-module adaptation chain. Second, a set of functional module agents is constructed, and a multi-agent reinforcement learning model is established to generate the optimal detection model structure under module priority and order constraints. Subsequently, the optimal detection model structure is encoded, a conditional generative adversarial network is constructed, and convolutional neural network parameters are generated based on the structure encoding vector and defect feature condition vectors. The optimal model parameters are then obtained through adversarial optimization. Finally, the optimal model parameters are loaded into the optimal detection model structure to generate the wafer image defect detection model. This invention achieves the collaborative generation of the detection model structure and parameters, improving the model's adaptability under different process conditions and multi-form mixed defect scenarios.

The invention discloses an ultra-high-performance concrete suction pile foundation structure of an offshore platform, which belongs to the technical field of ocean engineering and comprises a pile foundation main body, the pile foundation main body is formed by pouring ultra-high-performance concrete, a central interlayer is arranged in the side wall of the pile foundation main body, and a plurality of holes arranged in an array are circumferentially formed in the central interlayer. On one hand, the ultrahigh strength, high toughness and excellent durability of the ultrahigh-performance concrete material are fully utilized, and the problems that a traditional steel pile is prone to corrosion and frequent in maintenance are fundamentally solved; and on the other hand, material distribution is optimized through hole design, the ultrahigh pressure resistance and bending resistance of the pile foundation are guaranteed, and meanwhile the self weight of the structure is obviously reduced. According to the design, collaborative improvement of high strength, high durability and light weight is achieved, the structure is more complete, stress is more reasonable, the transportation, installation and full-life-cycle maintenance cost is greatly reduced, and a more efficient and reliable foundation support is provided for an ocean platform.

The present application relates to the technical field of carbon sink estimation, and discloses a mechanism-data driven coastal-estuarine ecocarbon sink evaluation method, system, terminal and storage medium, the method is around the idea of mechanism diagnosis-model construction-regional application, a mechanism constraint and data driven combination of coastal-estuarine ecosystem carbon sink estimation framework is constructed, a preset algorithm is used to obtain GPP sequence, Re sequence and NEP sequence, the key environmental driving law is identified by combining the generalized additive model, the GPP estimation model and Re response function are optimized, and the NEP estimation result is further coupled, the effective connection between empirical law identification and process parameterization expression is realized. The present application firstly performs mechanism identification driven by data, and then writes the key law back to the process expression, improves the model structure rationality, parameter interpretability and regional applicability, and improves the carbon sink evaluation accuracy of the coastal-estuarine ecological region.

The invention discloses a simulation test method for a dexterous hand. The simulation test method comprises the following steps: constructing a finite element simulation model based on the configuration of the dexterous hand; performing multi-physical-field multi-working-condition coupling simulation on the finite element simulation model to obtain simulation data; verifying the simulation data according to a pre-constructed bidirectional calibration mechanism and experimental data so as to correct the finite element simulation model; and when the corrected finite element simulation model meets a preset condition, outputting a dexterous hand performance evaluation result. By constructing a finite element simulation model, multi-physical field and multi-working-condition coupling simulation is carried out, so that the simulation model can output multi-dimensional performance data, and evaluation limitation is avoided. And meanwhile, aiming at a complex coupling environment in real operation, dynamic feedback optimization of the model is realized through a pre-constructed bidirectional calibration mechanism and experimental data verification, so that simulation data can be directly used for subsequent model iteration, and a stable and reliable dexterous hand simulation test process is realized.

A time-scale-based multi-modal semantic topology and constraint reasoning CAD sequence generation method belongs to the cross technical field of computer-aided design and artificial intelligence. The method constructs an end-to-end automatic generation framework from multi-modal input to editable parameterized CAD model, significantly improving the semantic consistency, structural rationality and logical correctness of the generated sequence; it can deeply integrate multi-source data such as images, sketches and point clouds, and accurately capture the design intent by using time-scale-aware feature extraction and dynamic topology graph evolution, and effectively overcome the error accumulation and logical drift problems in traditional autoregressive models through the sequence synthesis mechanism guided by syntax and geometric constraints, and realize efficient and robust generation of parameterized models combined with the geometric feedback-driven optimization engine, thereby achieving systematic breakthroughs in generation quality, automation level and cross-scene applicability, and providing key technical support for intelligent manufacturing and intelligent design.

This utility model discloses a uniformly vented dual-magnetic-circuit moving-coil loudspeaker, including a frame, a diaphragm voice coil mounted on the top of the frame, a partition assembly mounted on the bottom of the frame's inner cavity, and a resonant cavity on opposite sides of the frame and partition assembly. A partition vent hole is circumferentially formed on the partition assembly, and a frame vent hole is circumferentially formed at the bottom of the frame. This utility model has the advantage of uniform venting, solving the problem that existing moving-coil loudspeakers, during use, do not easily and smoothly expel air from the speaker's interior, and when the diaphragm's vibration amplitude is too large, it easily causes unbalanced diaphragm vibration due to impeded air expulsion.

A monocular image-based three-dimensional lane line detection method, device and medium, two-dimensional semantic features are extracted from a monocular lane image, and three-dimensional perception features are obtained by using a pre-trained deep model; two-dimensional semantic features and three-dimensional perception features are fused, an initial lane query is enhanced, and a query representation for three-dimensional lane line prediction is obtained; the query representation is iteratively decoded layer by layer, the sampling offset in the decoding process is dynamically adjusted based on the uncertainty parameters obtained by decoding, and the two-dimensional semantic sampling values and three-dimensional perception sampling values obtained by sampling in the decoding process are adaptively weighted and fused to obtain a three-dimensional lane line prediction result; the detection network is trained and constrained based on geometric structure constraints, and the final three-dimensional lane line detection result is output. The present application improves the spatial perception ability, prediction stability and lane structure rationality of monocular three-dimensional lane line detection in complex road scenes without relying on additional depth sensors and depth labeling information.

The embodiment of the invention provides a random map area generation method and device, equipment, a medium and a product, and the method comprises the steps: receiving a pre-configured map generation parameter, and determining a topological structure diagram corresponding to a target area based on the map generation parameter; mapping the topological structure diagram into a map plate corresponding to the target area to obtain a to-be-processed map dividing the target area into at least one sub-area; and adjusting the to-be-processed map based on the region identifier in the map generation parameter, the terrain type corresponding to the region identifier and the corresponding area distribution weight, and generating a target map matched with the map generation parameter. According to the technical scheme, the problem that the adjacency relation in random region generation is difficult to accurately control is solved, and meanwhile, the area error of the generated region is reduced.

The present application relates to the field of electronic technology, and discloses a method, device and equipment for underwater three-dimensional reconstruction of inspection well based on sonar ring scanning, the method comprising: obtaining sonar ring scanning data, attitude data of different depths of the inspection well and reference sonar ring scanning data of the inspection well; performing translational motion compensation processing on the sonar ring scanning data of each depth based on the attitude data of each depth and the reference sonar ring scanning data, to obtain first target data of the corresponding depth; extracting well wall points from the first target data of each depth, to obtain a first well wall point set of the corresponding depth; constructing a rotation matrix of the corresponding depth based on the attitude data of each depth; determining a second well wall point set of the corresponding depth in a global coordinate system based on the rotation matrix of each depth and the first well wall point set of the corresponding depth; and generating an underwater three-dimensional grid model of the inspection well based on the second well wall point sets of multiple depths, which can completely and intuitively restore the complex spatial topological form inside the inspection well.

The utility model relates to the technical field of electric appliances, in particular to a four-way connector, a drainage system, cleaning equipment and a cleaning system. The four-way connector comprises a main pipe, a first branch pipe and a second branch pipe. The main pipe comprises a first pipe opening, a second pipe opening and a third pipe opening, a separation part is arranged in the main pipe, and the first pipe opening and the third pipe opening are separated by the separation part; the first branch pipe is connected to the main pipe and is communicated with the third pipe orifice; the second branch pipe is connected to the main pipe and communicated with the third pipe opening. According to the technical scheme, the connecting structure of the pipeline system in the cleaning equipment is optimized, and the structural rationality of the cleaning equipment is improved.

This invention discloses a repository-level code translation agent method and apparatus based on a code graph structure, relating to the field of multi-agent technology. The method includes: constructing a repository-level code translation agent framework comprising a repository function understanding agent, a code translation agent, and a code inspection and repair agent; inputting the source Python code repository into the repository function understanding agent for functional analysis, establishing a correspondence between source Python code components and target Java language code files; determining the translation order of the target Java language code files based on the correspondence using the code translation agent; translating each target Java language code file sequentially to obtain translation results, constructing a target Java language repository; inputting the target Java language repository into the code inspection and repair agent, performing repairs using pre-configured processing tools, and outputting a complete and runnable target Java language repository. Using this invention can improve the completeness and consistency of repository-level code translation results.

The utility model discloses an integrated study desk and chair, comprising: chair subassembly has movable first base and fixedly set up on the chair body of first base, desk subassembly has movable second base and fixedly set up on the table body of second base, wherein, the edge of one of first base and second base is equipped with lock groove, and the other is equipped with locking mechanism, when first base and second base are close to each other, lock groove and locking mechanism are aligned and locked to limit first base and second base relative separation. The utility model provides an integrated study desk and chair, through setting up lock groove and locking mechanism of alignment locking between movable first base and second base, realize the quick disassembly of study desk and chair, promote structural stability, use flexibility and space adaptability, and it is convenient to carry, place and subsequent maintenance.

This invention proposes a segmentation-guided automatic region-level recognition method for rock thin section images. The method first performs target segmentation on the rock thin section image, obtaining multiple independent regions and corresponding region masks. Then, each independent region undergoes preprocessing including cropping, background suppression, boundary compensation, and size normalization to construct classification input samples. Next, the preprocessed region images are input into a recognition model to extract region texture, morphological, and semantic features, outputting corresponding category labels. Finally, the recognition results of each independent region are mapped back to the original image, and automatic analysis results are generated through region merging and consistency correction. This method effectively connects segmentation results with region-level recognition, improving the accuracy and automated analysis capabilities for identifying mineral grains, pores, and matrix targets in rock thin section images.

The present application relates to the field of radar image processing and target reconstruction, in particular to a single-view SAR three-dimensional target reconstruction method, comprising: training a three-dimensional generative prior model; constructing a compact latent space reparameterization network based on the model to convert original latent variables into low-dimensional compact latent variables; obtaining a single-view SAR observation image and imaging parameters of a target to be reconstructed and preprocessing; combining the three-dimensional generative prior model, the compact search latent space and a SAR physical forward imaging model to construct a physically consistent likelihood function, and using Markov chain Monte Carlo sampling for posterior inference to obtain a posterior compact latent variable sample set; determining a representative compact latent variable according to the sample set, generating a three-dimensional reconstruction result of the target, calculating three-dimensional structure uncertainty information of the target, and obtaining a confidence distribution and uncertainty evaluation result.