16results about How to "Guaranteed scalability" patented technology

The utility model relates to the technical field of aerospace and mechanical and electrical product assembly, in particular to a civil aircraft cable laying quality evaluation device, which comprises a shell, a bottom plate, a high-resolution camera and a switch, a CPU (central processing unit) processor and a battery are arranged in the shell, the bottom plate is arranged at the bottom of the shell, the high-resolution camera is mounted on one side of the bottom plate, and the switch is arranged at the left lower part of the shell. The front side and the rear side of the shell are each provided with a plurality of symmetrically-arranged heat dissipation holes. According to the utility model, the hidden hinged handle is linked with the clamping block, the compression spring and the limiting frame to realize one-hand operation adaptability, and the high-resolution camera and the processor are integrated to acquire cable point cloud data in real time and generate a three-dimensional model; the hinged dust cover is combined with diversified interfaces to guarantee interface protection and multi-equipment expansion capability, and the light supplementing lamp can be adaptively adjusted according to ambient light and is matched with the protective cover to improve the adaptability to complex illumination scenes; the symmetrical heat dissipation layout and the anti-skid cushion cooperatively suppress operation vibration, and the operation stability in a narrow space is ensured.

This application provides a software switch caching acceleration method and related apparatus based on FPGA, belonging to the field of network packet processing technology. The method includes: compiling a caching rule set into a decision tree forest using a decision tree construction algorithm, the decision tree forest comprising multiple decision trees of equal depth; mapping the decision tree forest to multiple parallel tree pipelines, the tree pipelines running on the FPGA, each tree pipeline corresponding to a decision tree, and each pipeline stage corresponding to a node layer of the decision tree; broadcasting a preset data packet to all tree pipelines for parallel processing; for each tree pipeline, guiding the preset data packet to traverse each pipeline stage until the preset data packet reaches the target leaf node of the decision tree; performing parallel rule matching on the preset data packet at the target leaf node of each decision tree to obtain the target caching rule matching the preset data packet, thereby accelerating flow table caching classification.

This invention discloses a smart building energy consumption optimization and comfort control system based on BIM-IoT integration, comprising: a BIM-IoT integrated digital twin module; a load prediction module based on a thermodynamic model; a multi-system collaborative optimization control module; and an equipment fault diagnosis module. By constructing a building digital twin model and integrating BIM static parameters and IoT dynamic data, this invention achieves accurate prediction of building heat load, multi-system collaborative optimization control, and intelligent balance between energy consumption and comfort, thereby significantly reducing building operating energy consumption and improving user experience.

The application discloses a kind of offshore wind turbine fault interval time prediction methods based on knowledge graph, comprising the following steps: step one: collection operation maintenance data generates sample set;Step two: time alignment and field standardization generate structured record;Step three: construct knowledge graph to obtain node set, edge set, edge weight;Step four: establish the mapping relationship of to-be-sampled parameter vector, generate graph atlas correlation matrix and construct structured proposal distribution;Step five: generate candidate parameter vector;Step six: divide subgraph data block and establish subgraph log likelihood cache, calculate log likelihood increment;Step seven: delay accept Metropolis-Hastings iteration update to-be-sampled parameter vector to obtain posterior sample set;Step eight: output fault risk parameter and generate fault interval time prediction result.The application improves the stability and calculation efficiency of fault interval time prediction result.

The invention provides an IVR (Interactive Voice Response) framework implementation method and system based on JAVA, a VXML 2.0 protocol is taken as a basis, a form of a custom XML (Extensible Markup Language) label is packaged through JAVA according to a service scene, and the custom XML label is packaged with a VXML 2.0 standard function and a service logic to realize various functions in an IVR service; in the telephone traffic request initialization scene, the IVR service platform triggers a service request through an HTTP interface request mode; in a business process assembly scene, arranging and assembling the packaged self-defined labels according to business scene requirements; and in the business analysis scene, the assembled XML document is output, and the IVR service platform carries out analysis operation to complete related telephone traffic actions, so that the rapid completion of the IVR telephone traffic process is realized. According to the invention, the service calling function of the IVR process can be realized in a relatively universal manner. And application developers only need to have a JAVA basis, so that the difficulty of application development is reduced, and the research and development quality is also ensured.

The application discloses a kind of calculation methods for simulating toxic and harmful gas diffusion concentration in three-dimensional virtual training environment, S1, determine toxic gas diffusion effective influencing factor;S2, the physical model corresponding to training environment is constructed, and the effective influencing factor of gas diffusion is derived;S3, according to gas toxic dose and shelter size, sparse grid is divided, and grid density is distinguished;S4, CFD model is constructed, and the concentration of simulated harmful gas diffusion is calculated;S5, concentration value time sequence situation result offline database of a period of time is established, and gas concentration situation result is stored;S6, when training, in virtual environment, quickly display.The method of the present application can consider terrain, building, meteorological conditions and other factors affecting gas concentration diffusion, can measure the gas diffusion concentration value of any point in space in real time, and its diffusion concentration value is realistic and practical, so that the simulation effect of harmful gas diffusion concentration value in virtual reality chemical hazard simulation training environment is more realistic and credible, and the training effect is improved.

This invention proposes a structure-condition-driven low-resource multilingual multimodal model training method and system, belonging to the field of artificial intelligence. The method includes: S1: Extracting image structural attributes from multimodal training samples, constructing structure-conditional text sequences, and obtaining text hidden states and visual features; S2: A route selector determines the training path based on the resource and structural attributes of the samples; S3: Freezing the multimodal large-scale model backbone network and dynamically activating a lightweight script adaptation module to update parameters; S4: Aligning visual and textual representations through a structure-conditional cross-attention module to generate a joint representation; S5: Embedding the joint representation, task prefix, target language identifier, and its language into the input decoding and multi-task output modules to obtain the generated result; S6: Constructing a total training objective function to obtain the trained multimodal model. This invention improves the overall performance, stability, and security of multimodal models in low-resource, multilingual, and complex structural scenarios.

The utility model discloses a power bank with a telescopic line structure, which relates to the technical field of power banks, and comprises a shell, the top end of the shell is covered with a top cover, one side of the shell is provided with a charging head and a charging line, the shell is internally provided with a storage disc, the storage disc is internally provided with a telescopic assembly, and the telescopic assembly is connected with the charging head and the charging line. The telescopic assembly comprises a center pipe, a lifting rod, a moving rod, a spring, a connecting line, a fixing clamping piece and a handle rope. According to the power bank with the telescopic line structure, the telescopic assembly is arranged, four springs are used for replacing a single metal elastic piece in the prior art, elastic stretching and retracting of the charging line are achieved, the service life of stretching and retracting of the charging line can be effectively prolonged, and the stretching and retracting effect can be guaranteed; according to the portable power bank, a user can conveniently carry the portable power bank to hook or carry the portable power bank, and the portable power bank can be pulled out for charging only by pulling the handle rope upwards to drive the four moving rods to retract inwards, so that the use functions of the portable power bank are enriched, and the use experience is improved.

The application discloses a high-sensitivity dielectric micro sensor suitable for monitoring salt migration of frozen soil and a monitoring system for salt migration of frozen soil. The sensor comprises a signal line, a probe and a probe needle, wherein one end of the signal line is electrically connected with an input end of the probe, and a measuring circuit is arranged in the probe and electrically connected with an electric connection end of the probe needle; the probe needle is two parallel slender conductors, the axes of the two probe needles are parallel to each other and fixed in the inside of an output end of the probe, keep a constant interval and are connected with the measuring circuit. The sensor is used for accurately detecting the dielectric constant of soil, and is directly related to the type / amount of salt under the conditions that the soil properties, compactness, water content and temperature are constant; even if the salt content changes slightly, the sensor can rapidly sense and instantaneously detect, and the dynamic monitoring and real-time display of the salt migration of the soil unit can be reliably realized by combining collection-transmission-processing-noise removal-demodulation-recording-screen display.

This invention discloses a joint deployment optimization method for hybrid expert model inference in heterogeneous GPU environments. The method first uses offline trajectory statistics to determine expert activation frequencies and collects resource profiles of heterogeneous devices. Then, it constructs a joint optimization model with the objective of minimizing the weighted sum of inference latency surrogate terms and quantization error penalty terms. Under memory constraints, a pruned greedy search algorithm is used to solve for the placement location, number, and quantization accuracy of each expert's replicas. During the online inference phase, a routing mechanism combining local priority and load-aware weighted round-robin is used for dynamic token distribution. This invention can simultaneously address inference latency, memory usage, and accuracy preservation in heterogeneous GPU environments, improving the processing efficiency of popular experts, reducing cross-device communication overhead and slow device trailing effects, and is suitable for distributed large-scale model inference deployment scenarios.

This invention provides a method and storage medium for 3D reconstruction of petrochemical pipelines based on multi-view optimization, relating to the fields of petrochemical pipeline inspection and 3D reconstruction. The method involves acquiring multiple pipeline images around the target petrochemical pipeline to construct a multi-view image set; preprocessing the multi-view image set; extracting features from each image to obtain a feature point set and feature descriptors; performing initial matching to obtain initial feature matching pairs; optimizing the process to obtain the optimal feature matching pairs; calibrating camera parameters to eliminate lens distortion and accumulated errors; generating a sparse point cloud using an improved motion recovery structure algorithm; removing redundant and noise points to obtain an optimized sparse point cloud; generating a dense point cloud using a dense reconstruction algorithm; optimizing the process to obtain an optimized dense point cloud; constructing an initial 3D model; and post-processing to obtain the final 3D model of the target petrochemical pipeline; and storing the data on a storage medium. This invention provides reliable 3D model support for the digital management, fault detection, maintenance, and repair of petrochemical pipelines.

The embodiment of the present application provides a scroll display device, comprising: a flexible display panel and a supporting shaft supporting the flexible display panel; the flexible display panel comprises a first end, a second end and a bending area connecting the first end and the second end, the bending area is curled along the outer wall of the supporting shaft; the first end and the second end can relatively move along a first direction with the rotation of the supporting shaft, and make the side, away from the supporting shaft, where the first end of the flexible display panel is located emit light; wherein the first direction intersects with the axial direction of the supporting shaft; the scroll display device further comprises: a first supporting strip located on the side of the second end facing the first end, the first supporting strip is used for cooperating with the supporting shaft to support the flexible display panel, so as to limit the shape of the second end or the position of the second end relative to the supporting shaft. The embodiment of the present application limits the shape and position of the flexible display panel by using the first supporting strip, avoids the flexible display panel from being out of the supporting shaft or from being deformed accidentally, and guarantees the display effect of the scroll display device.

This invention discloses a joint deployment optimization method for hybrid expert model inference in heterogeneous GPU environments. The method first uses offline trajectory statistics to determine expert activation frequencies and collects resource profiles of heterogeneous devices. Then, it constructs a joint optimization model with the objective of minimizing the weighted sum of inference latency surrogate terms and quantization error penalty terms. Under memory constraints, a pruned greedy search algorithm is used to solve for the placement location, number, and quantization accuracy of each expert's replicas. During the online inference phase, a routing mechanism combining local priority and load-aware weighted round-robin is used for dynamic token distribution. This invention can simultaneously address inference latency, memory usage, and accuracy preservation in heterogeneous GPU environments, improving the processing efficiency of popular experts, reducing cross-device communication overhead and slow device trailing effects, and is suitable for distributed large-scale model inference deployment scenarios.

The invention discloses a physical unclonable anti-counterfeiting material based on a DNA (Deoxyribose Nucleic Acid) nanostructure as well as a preparation method and application of the physical unclonable anti-counterfeiting material, and belongs to the technical field of information security and nano biology. According to the anti-counterfeiting material, three regular tetrahedron DNA monomers are connected through flexible double chains to form a DNA tripolymer, double-chain thermodynamic flexibility endows the anti-counterfeiting material with a nanoscale random geometric configuration, multi-dimensional encryption is achieved through different spectrum fluorophores modified randomly, and the information entropy reaches 4.3 bits / pixel. By combining deterministic self-assembly of the DNA nano-structure with dynamic flexibility, unification of deterministic preparation and random anti-counterfeiting characteristics is realized, and the method has excellent biocompatibility, high safety and strong expansibility, can be seamlessly compatible with a DNA computing system and a data storage system, and has wide application prospects. Multiple scenes such as information encryption authentication, biomedical authentication, flexible electronic anti-counterfeiting and DNA storage chip anti-counterfeiting are adapted, and an innovative platform is provided for the next generation of security authentication technology.

The application discloses an implementation method of a high-diffusion lightweight block cipher (HDLBC), which comprises the following steps: obtaining plaintext or ciphertext as to-be-encrypted / decrypted data; performing R-round encryption / decryption operation on the to-be-encrypted / decrypted data; each round of encryption operation is to perform RA1 function operation, branch XOR operation, RA2 function operation and P permutation on the to-be-encrypted data; in each round of encryption operation, part of bits in a sub-key participates in round key addition operation in the RA1 function operation; the remaining bits participate in round key addition operation in the RA2 function operation; the decryption operation is performed in the reverse order of the encryption operation, and the order of the sub-key used in the R-round decryption operation is opposite to the order of the sub-key used in the R-round encryption operation. The method designed in the application is an F function based on NAND operation and designed to introduce two F functions to realize round functions affecting all branches, so that the diffusion and area consumption of the algorithm are better ensured.

The application discloses an emotion-perceived music recommendation method and system based on electroencephalogram signals and a large language model, and belongs to the technical field of electroencephalogram signal processing and intelligent recommendation. User electroencephalogram signals are collected through a portable device, and after filtering and denoising, multi-dimensional features are extracted to identify the current emotional state; a trinity structured prompt of 'emotion-user-scene' is constructed in combination with user personal information, historical preferences and real-time environmental data; a target parameter package containing music style, emotional label, BPM range and instrument preference is generated by inputting a large language model for reasoning; candidate music is retrieved from a pre-labeled music knowledge base accordingly, and no more than 10 high-quality candidates are rearranged and screened in combination with a user portrait; finally, the top 3 music with scores output by a weighted scoring model are used as the recommendation result. The application combines physiological signals and multi-dimensional context, improves the accuracy of emotion recognition and the level of personalized recommendation, and is suitable for intelligent wearable devices and psychological health assistance.