9results about How to "Improve initial performance" patented technology

The application relates to a kind of distributed unmanned aerial vehicle ad hoc network routing methods based on deep reinforcement learning, comprising;According to Markov decision process, the deep reinforcement learning architecture of unmanned aerial vehicle communication network is built;Running Dijkstra algorithm sends original data packet from source node to destination node and generates original training data according to the routing process of original data packet to pre-train the deep reinforcement learning architecture;The coordinates of the destination node D of target data packet are input, the next hop node B of current node A is obtained using the pre-trained deep reinforcement learning architecture, and target training data is generated, and the deep reinforcement learning architecture is retrained according to the original training data and target training data;Next hop node B is used as starting node, until the next hop node is the destination node, the routing of target data packet is completed, and the application can enhance the robustness of network and improve the life of unmanned aerial vehicle communication network.

This invention proposes a water level detection system and method based on a pre-trained model, comprising: an image acquisition unit, a data preprocessing unit, a pre-trained model storage and retrieval unit, a data comparison and local data generation unit, a manual annotation and screening unit, and a fine-tuning training and model update unit. Through a two-stage model optimization process of "pre-training and local fine-tuning," combined with intelligent deviation analysis and a closed-loop iteration mechanism, it overcomes the pain points of traditional water level detection such as low efficiency, poor model generalization, and uncontrollable data quality. It achieves a high-precision, highly adaptable, and low-maintenance water level monitoring solution with significant industry application value.

This invention discloses a highly bioavailable ticagrelor solid dispersion, its formulation, preparation method, and application. The solid dispersion is prepared by hot melt extrusion of ticagrelor and a specific ternary carrier system, wherein the ternary carrier system is from Soluplus. ® The formulation consists of copovidone and hydroxypropyl methylcellulose. This solid dispersion is further combined with disintegrants, fillers, and lubricants to form a pharmaceutical composition. This invention utilizes the synergistic effect of a ternary carrier and hot-melt extrusion technology to ensure ticagrelor exists stably in an amorphous form, significantly improving drug dissolution and oral bioavailability. Experiments show that the formulation exhibits a dissolution rate ≥85% in pH 6.8 phosphate buffer after 30 min, and a relative bioavailability in beagle dogs reaching 150%–250% of commercially available tablets. It also demonstrates excellent stability, with no crystallization observed in a 6-month accelerated assay, indicating significant clinical value and market potential.

The invention belongs to the technical field of electric meter measurement detection, and particularly relates to an electric energy meter aging process temperature and measurement accuracy anomaly detection method and system, and the method comprises the steps: obtaining source field data and target field data; performing field adaptive pre-training on the feature extractor and the error regression device according to a self-defined global loss function to obtain a pre-training model; generating a personalized fine tuning factor for the target ammeter according to the complexity, reflected by the target domain data, of the environment where the target ammeter is located; and on the basis of the personalized fine tuning factor, performing online fine tuning on the pre-training model deployed on the target ammeter to obtain a real error prediction value of the ammeter to be detected, and judging whether the aging process temperature or the metering accuracy of the ammeter is abnormal or not according to the real error prediction value. According to the invention, domain deviation can be eliminated from the pre-training stage, and adaptive online fine tuning is realized, so that the accuracy of metering error correction in the full life cycle of the electric meter is improved.

The application belongs to the technical field of power load prediction and the field of network flow prediction, discloses a power flow prediction method based on neural network and boosting, combines a GRU-Xgboost module, a residual adjustment module and a missing value reconstruction, carries out decomposition reconstruction and adjustment on a complex network flow time sequence, proposes a new network flow time sequence prediction model, considers time correlation of time sequence data, and improves prediction accuracy.For the difference of the power network, according to the service characteristics of the power, the characteristic factors influencing the flow model and the effective processing of the nonlinear complex characteristics and unknown noise of the flow sequence are found, so that the flow prediction model for the power network is established, and the prediction information is provided for the relevant departments to make decisions.

The application relates to the field of magnet materials, and particularly discloses a neodymium-iron-boron material and a preparation method and application thereof. The preparation method of the neodymium-iron-boron material comprises the following steps: S1, respectively melting and pouring main phase material and auxiliary phase material to obtain main phase rapid solidification tablets and auxiliary phase rapid solidification tablets; S2, respectively performing hydrogen breaking treatment on the main phase rapid solidification tablets and the auxiliary phase rapid solidification tablets to obtain main phase coarse powder and auxiliary phase coarse powder, and respectively performing airflow grinding to obtain main phase fine powder and auxiliary phase fine powder; S3, performing pressing on a mixture containing the main phase fine powder and the auxiliary phase fine powder to obtain a blank, and performing sintering to obtain a neodymium-iron-boron material matrix; and S4, performing diffusion treatment on the neodymium-iron-boron material matrix by using a diffusion source; the obtained neodymium-iron-boron material has excellent remanence and coercive force, and good stability of magnetic performance.

The invention discloses a bonding type UV viscosity-reducing adhesive tape as well as a preparation method and application thereof in wafer cutting, and the preparation method specifically comprises the following steps: (1) adding a thermal initiator into a photoinitiator monomer and an acrylate monomer, and carrying out free radical copolymerization reaction to obtain a matrix adhesive solution; (2) adding a photoactive monomer and an active tin reagent into the matrix glue solution, and carrying out grafting reaction to obtain a UV peelable glue solution; (3) coating the UV peelable glue solution on the surface of a base material layer to form a UV peelable glue layer; and (4) covering a release film on the surface of the UV peelable adhesive layer to obtain the bonding type UV viscosity-reducing adhesive tape. The bonding type UV viscosity-reducing adhesive tape prepared by the invention has the following characteristics: the initial peel strength is high, and a protected material can be firmly fixed; after UV irradiation, the peeling strength is obviously reduced, and the material is easy to peel off from the surface of the material; and the whole process is simple, and has the remarkable advantages of no pollution, no adhesive residue and the like.

The present application relates to the technical field of automatic driving, in particular to a depth map completion method based on double-branch gradual fusion, comprising: processing color images and laser radar sparse depth images of the same target scene through a double-branch depth completion network based on mixed attention fusion and gradual refinement; training the double-branch depth completion network based on mixed attention fusion and gradual refinement using a training set in a depth completion dataset, obtaining network weight parameters after each round of training, and verifying the network weight parameters through a verification set, selecting the network weight parameters with the smallest error as the optimal network weight parameters; and performing depth completion on the double-branch depth completion network using a test set and the optimal network weight parameters to obtain a target dense depth image. The present application optimizes feature processing by extracting, fusing and refining the features of sparse depth images and color images, and finally obtains a dense depth image with rich details and accurate depth values.

The application discloses a method for preparing SnO2 thin film in iodine vapor environment and application thereof, and belongs to the technical field of battery device preparation. The preparation method comprises the following steps: (1) dissolving SnCl2 in isopropyl alcohol and stirring in air to obtain a SnO2 precursor solution; (2) performing ultraviolet ozone treatment on a substrate, and rotating and coating the SnCl2 precursor solution on the obtained substrate; after the spin coating, the substrate is placed in a closed iodine vapor environment for standing, and heating annealing treatment is performed to obtain the SnO2 thin film. On the basis of the solution preparation method, the SnO2 thin film is innovatively obtained by using iodine vapor oxidation and low-temperature heating, the method for preparing the SnO2 thin film is simple, and the preparation cost is relatively low; the prepared perovskite thin film will not be decomposed, the service life of the device can be prolonged on the basis that the efficiency of the perovskite battery device is kept close to the initial performance of the control group.