786results about How to "Improve throughput" patented technology

The invention provides a low-altitude logistics multi-unmanned aerial vehicle cooperative distribution method based on deep reinforcement learning, and the method comprises the steps: obtaining a distribution task and multi-level dynamic airspace information, and distributing the distribution task to idle unmanned aerial vehicles according to a task distribution mechanism; based on the multi-level dynamic airspace information, an airspace scheduling mechanism is constructed, and the airspace scheduling mechanism comprises an airspace scheduling height level selector and a multi-unmanned aerial vehicle cooperative path-finding solver; the airspace scheduling height level selector determines the flight height level of the unmanned aerial vehicle according to the distribution task and the multi-level dynamic airspace information to obtain a height level decision; the multi-unmanned aerial vehicle cooperative path-finding solver performs cooperative path planning in a horizontal range at each height layer to obtain a path decision; and the unmanned aerial vehicle executes a distribution task according to the height layer decision and the path decision. According to the method provided by the invention, the high-throughput and low-delay delivery target is realized in a complex urban scene, and the operation risk caused by airspace conflicts is reduced.

The invention relates to an engineering bacterium of high-yield ornamental blue as well as a construction method and application of the engineering bacterium, and belongs to the technical field of natural dye synthesis by a biological method. The engineering bacterium capable of highly producing the ornamental blue, disclosed by the invention, expresses icd, bpsA, glnAY405F and gd hA; acnR, yggB, glsK, aceA and ldh are knocked out from the engineering bacteria. According to the application, yggB and aceA genes are knocked out, site-directed mutagenesis (Y405F) is carried out on a glnA gene, a glsK gene is replaced by a glnAY405F gene, an acnR gene is replaced by an icd gene, an ldh gene is replaced by a gdhA gene, and a bpsA gene is integrated, so that the formation of by-products such as lactate and succinic acid in a blue-viewing synthesis process can be blocked; the flow direction of a passage from citric acid to isocitric acid to alpha-ketoglutaric acid to glutamic acid to glutamine is improved, so that the concentration of glutamine in cells is improved, and the yield of the ornamental blue is improved.

The invention discloses a key value storage system indexing method oriented to an NVM-NVMe SSD (Non-Volatile Memory-Non-Volatile Memory Express Solid State Disk) hybrid architecture, which comprises the following steps of: constructing a heterogeneous storage architecture taking cold and hot data perception as a core driving mechanism, coordinating and managing two types of storage media, namely a non-volatile memory NVM and a solid state disk NVMe SSD, and realizing efficient identification, layered writing and dynamic migration of cold and hot data. According to the method, the characteristics of low delay, durability and byte addressing of the NVM are utilized, the frequency and I / O overhead of Flush and Compaction operations are reduced in a data write-in path, meanwhile, the problem of mixed storage of cold and hot data is avoided, and therefore the overall performance and storage efficiency of a system are remarkably improved. In addition, by introducing an asynchronous migration module, the method can dynamically adapt to the change of data popularity along with time evolution, and effectively support high performance and high availability of the key value system in long-term operation.

The invention provides a Kyber and SM3 fused post quantum cryptography encryption system, method and device. The control logic module controls the execution round number of SM3 in the Hash module according to the random number length required by the NTT module in different modes; the interception unit intercepts the effective length of each round of hash value; the filling unit fills the input message of each round of SM3; the Hash module determines the number of rounds required to be executed by the SM3 according to the lengths of the hash values required by different modes, and intercepts the effective length of each round of hash value; the NTT module is used for calculating polynomial multiplication in the Kyber algorithm; the NTT module comprises a butterfly unit and a control unit, the butterfly unit is used for calculation of modular addition, modular subtraction and modular reduction, and the control unit determines an operation mode of the NTT module; and the sampling module comprises a sampling rejection unit and a central binomial distribution sampling unit, the sampling rejection unit is used for sampling to obtain a public matrix, and the central binomial distribution sampling unit is used for sampling to obtain a noise vector. The scheme meets the safety standard of China and is reliable in safety.

The invention provides a phase detection regulation and control system for an ultra-narrow-band Ku-band signal, and the system comprises a PL end which is used for carrying out the preprocessing of a plurality of paths of input Ku-band signals; the PS end is used for detecting the phases of the multiple paths of final baseband signals obtained through preprocessing to obtain a phase detection result and sending the phase detection result to the upper computer; the upper computer is used for generating a phase regulation and control instruction according to the phase detection result; and the phase shift control module adjusts the phase of the Ku wave band signal of the corresponding path. According to the method, multiple paths of narrowband signals can be synchronously extracted, phases of the narrowband signals can be analyzed, microsecond-level detection efficiency can be achieved, the LMS algorithm is introduced in the PL end processing process to extract narrowband target signals, a self-adaptive step length adjusting filter is introduced in the LMS algorithm, and detection precision is improved. The method has higher flexibility and adaptability, can effectively meet the processing requirements of different signal types in a complex environment, and improves the universality and applicability of a signal detection system.

The invention provides a continuous flow magnetic separation chip and a preparation method thereof.The chip comprises an upper pressing plate, a lower pressing plate, a chip bottom plate and a flow channel cover plate, the flow channel cover plate is arranged on the chip bottom plate in an attached mode, a flow channel is formed between the flow channel cover plate and the chip bottom plate, and the flow channel cover plate is further provided with a flow channel inlet and a flow channel outlet; the chip bottom plate and the runner cover plate are fixedly clamped between the upper pressing plate and the lower pressing plate, soft magnetic strip materials are packaged in the chip bottom plate and the runner cover plate, and the soft magnetic strip materials are located on one side of the runner. The soft magnetic strip material patterned array is used, and the soft magnetic strip materials are packaged in the chip bottom plate and the flow channel cover plate, so that double-layer arrangement is formed in the vertical direction, magnetic force distribution with the inner side direction of the flow channel as the main direction and the longitudinal direction as the auxiliary direction is achieved, the cell wall attaching effect is effectively weakened, high-flux cell magnetic control screening can be achieved, and the cell magnetic control screening efficiency is improved. And the magnetic field screening efficiency is improved. In addition, during cell sorting, gradient grading regulation and control can be achieved in the aspect of magnetic field force.

The invention discloses a deep mutation scanning method based on protein-protein interaction in bacterial cells and a binding affinity prediction method. In the invention, split dCas9 proteins (SEQ ID NO.1 and SEQ ID NO.2) are respectively subjected to fusion expression with a receptor and a ligand with binding affinity to be detected, and are combined with a toxic sgRNA guide sequence (SEQ ID NO.3), and when PPI occurs, a toxic dCas9: sgRNA complex is formed to inhibit bacterial growth, so that a PPI affinity size signal is converted into a bacterial cell generation speed signal, and the signal can be quantified into a copy number value through next-generation sequencing. The invention further designs a deep learning binding affinity prediction model, high-affinity sequences can be screened from theoretical mutation space, the workload and cost of experimental verification are remarkably reduced, and a reliable basis is provided for subsequent protein function optimization and design.

The invention provides a lightweight large model operation method based on end-side deployment. The lightweight large model operation method comprises the following steps: acquiring equipment operation data and storing a model file in a mixed precision format according to quantization precision supported by terminal equipment; determining a mixing precision quantification strategy and carrying out strategy analysis on the equipment operation data; a Key-Value cache file obtained through model reasoning in the historical dialogue is reserved; calculating a semantic embedding vector of the second user request, calculating a cosine similarity between the semantic embedding vector and the first V vector, and judging whether the second user request has a similar intention or not according to a calculation result; and if yes, performing incremental reasoning on the second user request by multiplexing the Key-Value cache file. By monitoring the state of the terminal equipment in real time, dynamically adjusting the quantization level, multiplexing the cache data of the semantic related requests and merging batch request reasoning, the effects of balancing the energy consumption and performance of the terminal equipment, reducing repeated calculation and improving the service time of the equipment and the user experience are achieved.

The invention provides an uplink joint receiving method and device and a storage medium. The method comprises the steps that first network equipment determines ephemeris link delay inequality at the current moment when receiving an uplink data signal sent by a terminal; the ephemeris link delay inequality is used for representing a delay inequality value between a first ephemeris link from the terminal to the first network equipment and a second ephemeris link from the terminal to the second network equipment; based on the ephemeris link delay inequality at the current moment and the ephemeris link delay inequality error value at the current moment, determining a fast Fourier transform (FFT) window time offset adjustment amount of the second network equipment; and sending the FFT window time offset adjustment amount to the second network device, wherein the FFT window time offset adjustment amount is used by the second network device and the first network device to jointly receive uplink data signals sent by the terminal.

The invention discloses an SNP (Single Nucleotide Polymorphism) molecular marker related to sheep fecundity and application of the SNP molecular marker. The SNP molecular marker comprises an SNP1 (Single Nucleotide Polymorphism) and / or an SNP2; the SNP1 is located at a base at the 115747587 site of a chromosome 6 of a sheep reference genome AR-UIRamV2.0 version, and the polymorphism of the SNP1 is T / G; the SNP2 is located at a basic group at the 41419566th site of a chromosome 22 of a sheep reference genome AR-UIRamV2.0 version, and the polymorphism of the SNP2 is A / G. According to the scheme, the SNP molecular marker related to sheep fecundity is provided, sheep with the high fecundity character can be efficiently and accurately identified and screened out, and the marker is high in typing quality, single in copy and high in polymorphism and can be used for sheep fecundity molecular marker assisted breeding.

The invention discloses a polyvinylidene fluoride (PVDF) super-hydrophobic membrane and a preparation method thereof.The preparation method comprises the steps that firstly, methyl sodium silicate, perfluorodecyl triethoxysilane (PFDTES) and tetraethoxysilane (TEOS) serve as catalysts / reactants, fluorinated hydrophobic modification is conducted on hydrophilic silicon dioxide (SiO2) nanoparticles, a nanoparticle non-solvent composite spraying synergistic phase inversion method is provided, and the polyvinylidene fluoride (PVDF) super-hydrophobic membrane is prepared. F-SiO2 is successfully introduced into the surface of the membrane, a multi-scale micro-nano structure is formed on the surface of the membrane, and the hydrophobicity and the surface roughness of the membrane are greatly improved. The prepared membrane is applied to vacuum membrane distillation testing, the flux and the salt rejection rate are very excellent, the maximum salt rejection rate reaches 99.99%, the flux is kept at 18.21 kg.m <-2 >. H <-1 > or above, the performance of the super-hydrophobic membrane is superior to that of a base membrane and a membrane prepared through a traditional non-solvent induced phase separation method, and the super-hydrophobic membrane has good anti-pollution performance and is very good in recovery effect after being washed with water.

The invention relates to the technical field of microfluidics and digital PCR detection, and discloses a microfluidic chip for centrifugal droplet generation and a digital PCR detection method.The microfluidic chip comprises a sealed cavity formed by a bottom plate and a cover plate, a through hole is formed in the center of the chip and used for being fixed to centrifugal equipment, the core structure of the chip is a droplet generation component, and the droplet generation component is arranged in the cavity. Comprising a sample adding cavity, a micro-nozzle array and a liquid drop collecting cavity, a sample is driven by centrifugal force to generate nano-liter-scale'water-in-oil 'liquid drops through micro-nozzles, the chip is used for digital PCR reaction, each liquid drop serves as an independent reaction unit to improve detection sensitivity and precision, the chip is suitable for detection of low-abundance gene mutation, and by combining Taqman-MGB probe or other molecule detection technologies, the detection sensitivity and accuracy are improved. The chip can efficiently detect mutation types of EGFR genes, including EGFR-T790M mutation, has a wide clinical application prospect, provides a high-throughput and high-precision droplet generation method, and is suitable for application scenarios of digital PCR, single cell analysis and other microfluidic technologies.

The invention discloses a data packet spliced inter-chip data transmission method and device and a storage medium, and the method comprises the steps: analyzing to-be-transmitted network data, obtaining to-be-transmitted data packets, and mapping the to-be-transmitted data packets to a plurality of data flow queues; constructing an initial DMA chip to carry out packet splicing processing on the to-be-transmitted data packets in the data flow queue to obtain a chained DMA chip; if the cache length of the chained DMA chip exceeds a preset cache length or the current packet splicing time exceeds a preset waiting time, transmitting the chained DMA chip to a next chip; and if the chained DMA chip is transmitted to an exit chip, obtaining all the data packets to be transmitted from the chained DMA chip and sending the data packets. According to the method, the multiple data packets are combined into one chained DMA chip for transmission through packet splicing processing, the overall transmission efficiency can be improved, and efficient transmission of a large amount of data among chips is ensured by dynamically adjusting the size and the transmission strategy of the chained DMA chip.

The invention discloses an internet-based online gynecological disease intelligent consultation system, and relates to the field of data processing, and the system comprises the steps: generating a public and private key pair of a patient through employing an RSA asymmetric encryption method; a BERT model based on deep learning is adopted to process the decrypted inquiry information, processing is carried out on local equipment, and structured patient condition description is obtained; the encrypted patient condition description is uploaded to a server, a system private key is used for decryption on the server, diagnosis is carried out by using a reasoning method based on a knowledge graph according to the patient condition description, and a disease candidate list is obtained; transmitting the encrypted candidate list to local equipment of the patient, decrypting by using a private key of the patient, and reasoning by adopting a CART decision tree according to the disease candidate list to obtain a final diagnosis result; aiming at the low precision of online gynecological disease consultation diagnosis in the prior art, asymmetric encryption, knowledge graph reasoning, a CART decision tree and the like are adopted, so that the diagnosis precision is improved while the privacy of a patient is protected.

An integrated cleaning and liquid supply device and online cleaning method for polyurethane polishing pads are disclosed. The device includes a base, a rotating seat, a swing arm, an independent liquid supply pipeline system and cleaning pipeline, a high-pressure nozzle array, and a positioning sensor assembly. During the polishing stage, polishing liquid is dripped through the liquid supply outlet. During the cleaning stage, the swing arm moves centripetally to a preset position aligned with the center of the polishing pad. The positioning sensor triggers a high-pressure pump, causing pure water to be sprayed as an atomized jet along the diameter of the pad surface through the high-pressure nozzle array. Simultaneously, the polishing pad rotates at a low speed, utilizing the synergistic effect of centrifugal force and high-pressure water flow to thoroughly remove residual contaminants within the micropores of the polyurethane. This invention integrates liquid supply and deep cleaning functions into the same motion mechanism, eliminating the need for machine shutdown or manual intervention. It offers advantages such as non-contact, non-destructive, thorough cleaning, and high automation, extending the life of the polishing pad and improving the stability of the CMP process.

The invention discloses a large model reasoning hardware accelerator based on data flow execution, and belongs to the technical field of large model reasoning hardware acceleration. Each node comprises a linear calculation kernel used for sequentially completing query / key / value projection and attention output projection in multi-head attention calculation, dimension raising projection and dimension reduction projection calculation in a feedforward neural network layer, and inverse quantization operation after each projection is output; the multi-head attention calculation kernel is used for completing loading and quantization of query / key / value vectors, multi-head division and cache management, attention score calculation, Softmax normalization, context vector generation and KV cache updating; the RLA kernel is used for completing residual addition, layer normalization, quantization / inverse quantization and nonlinear activation calculation of each layer; the three cores in the same node are located in the same SLR area, and data transmission is carried out through a hardware queue and an HLS blocking mechanism. According to the invention, high hardware utilization rate and low-delay reasoning can be realized.

The invention belongs to the technical field of block chains, and particularly relates to a financial and tax data fusion and audit traceability management system based on a block chain, which comprises an under-chain multi-source data acquisition layer, an on-chain dynamic fusion verification layer and a cross-domain collaborative feedback control layer, and is pre-bound to a data source through a business rule engine. Through combination of a lightweight fusion unit of space-time alignment and an incremental consensus mechanism based on a state channel, semantic consistency, fine-grained tracing and real-time compliance intervention of finance and tax data in a full life cycle are realized. Through a business rule pre-binding and source event capturing mechanism, finance and taxation data is prevented from breaking away from compliance context in a generation stage; a space-time alignment fusion unit is utilized to construct a composite transaction object with causal logic, and semantic segmentation among multi-source data is eliminated; the incremental consensus architecture based on the state channel is realized, and the final consistency of key transactions is ensured while the high throughput performance is maintained.

The invention discloses a block chain cross-chain transaction method, which is characterized in that a cross-chain transaction between two or more block chains by a user follows an optimistic execution mechanism, the transaction is executed by an executor under the chain, a zero-knowledge proof does not need to be generated for each block, the zero-knowledge proof is generated only when the transaction is challenged, and a transaction execution result is stored on the chain; when the transaction is challenged, the smart contract on the chain performs cross-chain transaction verification, specifically including execution verification and inconsistency verification, the method provided by the invention verifies the transaction itself, which belongs to direct verification, and when the transaction verification is not passed, the account state of the associated account on the chain is withdrawn until the account state is updated, so that the security of system transaction execution is ensured; according to the transaction method provided by the invention, an optimistic execution mechanism is adopted, so that the transaction does not need to be executed on a chain, and compared with the prior art, the transaction method has higher throughput and stronger expandability.

The invention discloses an operating system IO processing method supporting random IO merging, an operating system and a storage system, and belongs to the field of computer operating systems. The method comprises the steps that in a user mode, only one parameter copy is reserved for the same parameters of a plurality of random IO requests and recorded in a shared structure A, respectively recording different parameters of each random IO request into the shared structure body B to obtain a pseudo sequence request; the pseudo-sequence requests are issued to a system kernel, corresponding operations are executed on the pseudo-sequence requests in all layers of the system kernel in sequence, the batch operations are only executed once, and the non-batch operations are executed once for all the random IO requests respectively; the batch operation and the non-batch operation are operations executed for parameters in the shared structure body A and the shared structure body B respectively; and splitting the pseudo sequence request into random IO requests in an equipment driving layer of the kernel system, generating a corresponding equipment command, and submitting the equipment command to storage equipment. According to the invention, kernel overhead under the condition of high throughput can be relieved, so that the throughput is improved.

The invention discloses a liquid drop generation structure, a micro-fluidic channel, a micro-fluidic chip and a liquid drop generation method. The liquid drop generating structure comprises a side wall, a deformable part and a driving part, the deformable part is connected with the side wall to form a channel, and the deformable part is used for controlling opening and closing of the channel through deformation of the deformable part; the driving part is connected with the deformable part and used for driving the deformable part to deform. The microfluidic flow channel comprises the liquid drop generation structure. The micro-fluidic chip comprises the micro-fluidic flow channel. The droplet generation method comprises the following steps: injecting liquid into the flow channel body through an inlet of the flow channel body; the deformable part is driven by the driving part to deform so as to close the channel, liquid in the flow channel body extrudes the deformable part under the action of driving force so that a thin seam channel can be formed between the deformable part and the side wall, and the liquid passes through the thin seam channel to form jet flow liquid drops to be jetted out.

The invention provides an SM4 algorithm rapid implementation method and system based on a hardware extension instruction set, and the method comprises the steps: inputting a key according to 128 bits, and extending the key into a round key RK; performing data conversion on the round key RK to obtain a converted round key; the 16 input SM4 data blocks are cut and converted to obtain converted plaintext data; 32 rounds of iteration are carried out on the converted round key and plaintext data, an AVX512 instruction set instruction is used for achieving affine transformation to convert the data from a finite field where SM4 is located to a finite field where AES is located, then a VAES instruction set instruction is used for executing S box operation in the AES, and then the AVX512 instruction set instruction is used for achieving inverse affine transformation to convert the data from the finite field where the AES is located to the finite field where the SM4 is located; and performing large-end and small-end conversion on the data after the 32-round iteration process of the SM4 algorithm is executed again, taking out the data from the 512-bit register in a blocking manner, arranging the data into 16 blocks of ciphertexts corresponding to 16 blocks of plaintext data according to a specific sequence, and outputting the 16 blocks of ciphertexts.

The invention discloses a message processing task scheduling method and device which are applied to a distributed system, and the method comprises the steps: extracting the features of a to-be-processed message; analyzing the features of the message by using a convolutional neural network to obtain the complexity of a processing task of the message, the processing task being that a computing node in a distributed system asynchronously processes the message through a processing resource; based on the complexity of the processing tasks, priorities are allocated to the processing tasks, and the priorities of the processing tasks are in negative correlation with the complexity of the processing tasks; determining a scheduling strategy corresponding to the processing task based on the priority of the processing task; and executing the processing task according to the scheduling strategy.

The invention discloses a polyimide hollow fiber gas separation membrane and a preparation method thereof, and the polyimide hollow fiber gas separation membrane is prepared by reasonably designing a precursor structure and accurately regulating spinning parameters. The prepared polyimide hollow fiber gas separation membrane is compact in surface, free of holes and defects and excellent in mechanical property, and has a high-flux and high-selectivity gas separation effect, especially for hydrogen. Wherein the hollow fiber spinning solution is obtained by mixing a polyimide polymer, a high-boiling-point solvent, a low-boiling-point solvent and a non-solvent, and the polyimide polymer is synthesized by carrying out condensation polymerization on a dianhydride monomer, a diamine monomer containing a benzimidazole group and other diamine monomers. The preparation method provided by the invention is simple, convenient and easy to operate, avoids tedious procedures and additional economic investment brought by subsequent processes (polydimethylsiloxane coating and crosslinking modification), and has a wide application prospect.

The invention provides a wave position dynamic transfer method, system and device based on inter-satellite linkage and a medium, and the method is executed by an overload satellite and comprises the steps: judging whether a self load is overloaded or not; if the load is judged to be overloaded, selecting a to-be-transferred wave position from self-managed wave positions; acquiring state information of the adjacent satellites through inter-satellite interaction with the adjacent satellites; based on the obtained state information, determining a target satellite capable of bearing a to-be-transferred wave position from adjacent satellites having coverage overlapping with the target satellite; negotiating with the target satellite through an inter-satellite link to confirm beam position transfer, and triggering the target satellite to adjust the beam to cover the to-be-transferred beam position; and after negotiation confirmation, controlling the user equipment in the to-be-transferred wave position to be switched to the target satellite, and releasing the wave position resources. By adopting the method, the communication delay of the user equipment can be reduced, the throughput of the user equipment and the constellation resource utilization rate can be improved, and the service quality and user experience of a satellite communication system can be comprehensively optimized.

The application belongs to mass spectrometry technology, and specifically provides a novel electrospray ionization device and method. The ionization device comprises a power supply and an electrode. The electrode has a tip, and a sample solution is suitable to be attached to the tip. A voltage V applied to the electrode satisfies: k is a coefficient, α is an angle of the tip, ε0 is a vacuum dielectric constant, η is a viscosity of the sample solution, γ is a surface tension of the sample solution, d is a distance between an electrical access point on the electrode and the tip, and Q is a volume of the sample solution. The application has the advantages of no need for pretreatment and the like, and can be applied to on-site mass spectrometry detection.

The present application belongs to the technical field of wireless communication, low-altitude communication and artificial intelligence, and particularly relates to a high energy efficiency spectrum management method for a low-altitude communication system. The present application constructs a layered feature extraction mechanism for interference cognition, and progressively fuses link connection state, same-subchannel multiplexing state and user-side interference perception state according to the action path of spectrum interaction relationship, so that the obtained access point features and user equipment features can simultaneously contain connection quality, spectrum multiplexing, interference propagation and energy consumption state information. On this basis, the present application establishes a decision mapping mechanism for hybrid spectrum management actions, converts the aggregated features into subchannel allocation, user association and power allocation decisions, and jointly optimizes the feature extraction mechanism and the decision mapping mechanism through global energy efficiency feedback, thereby improving the spectrum management effect and global energy efficiency in a dynamic low-altitude communication system.

The invention provides an electrochemical method, an electrode and a system for synthesizing hydrogen peroxide without aeration, and relates to the technical field of electrochemical preparation of hydrogen peroxide. The multistage hole penetrating type cathode with a specific multistage hole channel structure is adopted, the oxygen-containing electrolyte penetrates through the multistage hole penetrating type cathode for electrochemical reaction under the non-aeration condition, dissolved oxygen in fluid is forced to make efficient contact with catalytic active sites through the multistage hole channel structure, the mass transfer process is greatly enhanced, and the catalytic activity of the catalyst is improved. And the diffusion limitation of the traditional flow-through electrode is overcome. And meanwhile, the structure required for synthesizing hydrogen peroxide is simplified by an aeration-free operation mode, and the energy consumption and the cost are reduced. Finally, through the synergistic effect of the technical characteristics, the high-yield and high-flux synchronous output of hydrogen peroxide can be realized under the condition that an expensive catalyst and external gas aeration are not needed.

The invention discloses a surface-modified composite fiber filter membrane as well as a preparation method and application thereof, and relates to the field of insulating oil filtration treatment. The preparation method comprises the following steps: uniformly stirring PVA powder, a solvent and ZrO2 nanoparticles to obtain a spinning solution; the spinning solution is prepared into the composite fiber filter membrane by adopting an electrostatic spinning technology, and the composite fiber filter membrane can be applied to the field of filtration and purification treatment of mineral insulating oil. The electrostatic spinning technology is adopted for preparing the filter membrane, the filter membrane has the advantages of being small in fiber diameter, good in uniformity, high in porosity and large in flux, ZrO2 nano particles are introduced into the filter membrane to increase the specific surface area and provide the impurity adsorption effect, trace moisture and micro impurities in the aged mineral insulating oil are adsorbed and intercepted, and the filter membrane has the advantages of being small in fiber diameter, good in uniformity, high in porosity and large in flux. Accurate filtering of the aged mineral insulating oil is achieved, the breakdown voltage of the aged insulating oil is effectively improved, and resource waste of the mineral insulating oil and pollution to the environment are reduced.

This invention belongs to the field of intelligent agent large language model service technology, and discloses a chip computing system, device and application for intelligent agent large language model. Through the collaborative design of hardware architecture and software scheduling, it systematically solves the core technical challenges faced by intelligent agent large language model services. Specifically, it includes a hardware and software collaborative implementation scheme for resource partitioning, key-value cache management and runtime scheduling of pre-filling, decoding and tool call related workloads in a multi-chip encapsulation and inconsistent memory access environment.

This invention discloses a communication method and system for anti-interference wireless ad hoc networks, relating to the field of wireless networking technology. The method includes: a master node collecting historical interference feature sequences transmitted by each slave node (not part of the ad hoc network) and predicting predicted interference frequency band sequences that may affect each node within a preset future time range; analyzing and quantifying the interference confidence of each predicted interference frequency band sequence by combining historical transmission interaction data between nodes; using the interference confidence as weight, jointly and dynamically allocating transmission time slot frequency bands of the ad hoc network within the preset future time range using an iterative optimization algorithm to generate multiple optimized time slot frequency band sequences capable of actively avoiding high-confidence predicted interference; and precisely controlling the communication behavior of each node within the future time period based on the multiple optimized time slot frequency band sequences. This invention transforms traditional passive interference response into active prediction and avoidance, improving the communication reliability and overall performance of wireless ad hoc networks in complex interference environments.