13results about How to "Insufficient relief" patented technology

The application discloses a shock-absorbing type culvert with animal migration and water discharge composite functions, and belongs to the technical field of animal passageways, which comprises a square culvert, shock-absorbing structures are arranged on the left side, the right side and the upper side outside of the culvert, the shock-absorbing structures relieve the vibration caused by vehicle driving, reduce the noise size in the culvert and prevent animals from being frightened. A composite light-emitting coating structure is arranged on the inside of the upper side of the culvert, the composite light-emitting coating structure weakens the light difference in the daytime and plays the night light performance at night, thereby providing guidance for animal migration. Drainage structures are arranged on the lower sides of the culvert, the drainage structures prevent the culvert from being washed away, animal passageways are arranged at the middle positions, outer extension plates are further arranged at the entrances and exits of the culvert, organic soil layers and water storage layers are arranged in the animal passageways, the outer extension plates and the organic soil layers and the water storage layers form a boat-shaped structure, water bodies are stored in the boat-shaped structure, the boat-shaped structure is beneficial to relieving the shortage of water resources in arid and semi-arid areas, the culvert is filled by plant growth, the culvert is closer to nature, the wariness of migratory animals to the culvert is reduced, and the ecological improvement effect is achieved.

This invention relates to a method and system for anomaly identification in power control terminals based on contrastive learning, belonging to the field of network data security technology. It solves the problem that existing technologies cannot effectively mine complex abnormal behaviors in power control data. The method includes: dividing network traffic at the data interaction interface into multiple time sliding windows, extracting multi-dimensional traffic features to obtain data interaction feature vectors; constructing a time-series feature graph using the data interaction feature vectors and constructing positive and negative contrast samples, and performing contrastive learning training on a graph attention network model; generating an embedding vector for each sliding window using the encoder of the pre-trained graph attention network model, scoring all sliding windows for anomalies based on the embedding vectors, and identifying sliding windows with abnormal data interaction behavior. This invention combines graph learning, contrastive learning, and time-series modeling techniques to provide a high-precision, low-dependency, and highly generalizable anomaly identification and control method.

This application provides a method for enriching lithium from low-grade lithium clay ore. The method includes the following steps: obtaining low-grade lithium clay ore; crushing and roasting the low-grade lithium clay ore to obtain sinter; scrubbing the sinter to cause it to rub against itself; performing secondary screening on the scrubbed sinter to obtain a primary screening material with a particle size less than or equal to 0.83 mm and a secondary screening material with a particle size less than or equal to 0.015 mm; and performing flotation on the primary screening material to obtain flotated material, wherein the flotated material is combined with the secondary screening material to form lithium concentrate. This method for enriching lithium from low-grade lithium clay ore can efficiently develop and utilize lithium resources in low-grade lithium clay ore.

The application discloses a flue gas-waste steam residual heat recovery large-temperature-difference heat supply system and method for power plants and heat source plants, solves the problem of low energy utilization efficiency in the prior art, has the beneficial effects of saving water and electricity and realizing full utilization of energy, and specifically has the following arrangement: the flue gas-waste steam residual heat recovery large-temperature-difference heat supply system for power plants and heat source plants comprises a flue gas residual heat recovery unit, a condenser, and a heat supply boiler; the flue gas residual heat recovery unit comprises a first heat exchanger, and treated flue gas enters the flue gas residual heat recovery unit for residual heat recovery; the waste steam residual heat recovery unit comprises the condenser; the condenser is connected with the heat supply boiler; the flue gas residual heat recovery unit is connected in series with the heat supply boiler; the heat supply unit comprises a first water supply pipeline, the heat supply boiler and the waste steam residual heat recovery unit are connected in series with the first water supply pipeline, a first branch of a second water return pipeline is heat-exchanged with the first water supply pipeline through a second heat exchanger, and the second heat exchanger is provided with the first water return pipeline and a second water supply pipeline.

The application discloses a steel surface defect image generation algorithm based on coordinate attention and multi-scale texture edge guidance, which comprises the following steps: S1, obtaining a steel surface defect image dataset, and performing a pretreatment operation on image data in the dataset; S2, based on the steel surface defect image dataset pretreated in step S1, preparing a diffusion model training dataset; S3, constructing a diffusion image enhancement model based on multi-scale texture-edge guidance, wherein the diffusion image enhancement model comprises a forward diffusion process, a noise prediction network and a reverse denoising generation process, and a multi-scale texture-edge structure enhancement module and a coordinate attention module are introduced into the noise prediction network; S4, based on the training dataset obtained in step S2, training the diffusion image enhancement model to obtain a noise prediction network for reverse denoising; S5, based on the trained noise prediction network, obtaining an enhanced steel surface defect image by using a structure-guided img2img defect generation mode; and S6, merging the enhanced steel surface defect image with the original steel surface defect image, constructing an expanded steel surface defect detection dataset, and performing experiments and evaluation on the enhancement effect. The application can enhance the defect edge structure expression while maintaining the continuity of the steel surface background texture, generate steel surface defect images with clear texture, reasonable structure and diversity, thereby effectively expanding the existing steel surface defect detection dataset and improving the training effect of the subsequent defect detection model.

The application discloses an optical crystal production dynamic optimization control system based on defect monitoring, which comprises a defect monitoring module, a reinforcement learning decision module, an in-situ repair execution module, a process optimization module, a data feedback module, a precision correction module and a federal learning collaboration module. A double-layer architecture combining inner real-time closed loop and outer cross-domain collaborative closed loop is adopted: the inner closed loop realizes millisecond-level real-time optimization through defect monitoring, real-time decision, repair execution and process optimization, and generates correction factors according to the correction factors generated by the data feedback and precision correction modules; the outer closed loop aggregates cross-batch, cross-process and supply chain data through the federal learning collaboration module to generate global optimization parameters and collaboratively enhance the decision. The reinforcement learning decision module synchronously receives and fuses the precision correction factors and the global optimization parameters, dynamically corrects the decision instructions, and the application realizes dynamic optimization control of the whole production chain.

The application discloses an ancient Chinese automatic sentence breaking and semantic analysis system based on deep learning, and relates to the technical field of ancient Chinese analysis.The system combines a gating recurrent unit (GRU) and a convolutional neural network (CNN) through a sentence breaking module, combines image character features, and can effectively process semantic ambiguity, long-distance dependence, and problems of homophones and variant characters.A knowledge graph containing historical and cultural common knowledge and dynamic relationships between words is constructed through a semantic analysis module, prior knowledge is injected through a graph neural network, and the depth of semantic understanding is improved.A sentence breaking result containing sentence breaking basis and a visual semantic analysis result are provided through an output module, so that users can clearly understand the processing logic and the trustworthiness and convenience are improved.The system can efficiently process a large number of unlabeled ancient Chinese texts, fully excavate the value of various cultural heritage such as classics and inscriptions, promote the research progress of grammar and vocabulary, and help the inheritance and spread of ancient Chinese culture.

The application discloses a robust micro-expression segment segmentation method and system which avoids head movement, adopts computer vision tools to recognize head posture of each frame image in a video sample, and extracts translation variable and rotation variable of the head posture relative to the same coordinate axis; adopts frame difference method to calculate head movement amplitude, divides the video sample into several continuous frame intervals with head movement amplitude less than a set displacement threshold; adjusts and aligns each frame of the video sample with continuous frame intervals to obtain a micro-expression sample; the system comprises a video sample library, a vision tool, a calculation module, a comparison module, a video segmentation module, a micro-expression processing module and a micro-expression sample library; the application divides small-amplitude head movement video segments which can be used as micro-expression samples, avoids the influence of head posture on micro-expression detection, eliminates the influence of head movement within a certain displacement threshold by alignment or removal of global motion, and improves the performance of micro-expression detection.

The invention discloses a container lockhole detection method based on HSLD-YOLO, and the method comprises the steps: obtaining a container lockhole image, constructing a lockhole small target training set and a test set, carrying out the preprocessing and enhancement of the container lockhole image, inputting the container lockhole image into a Backbone network, extracting a multi-scale feature map, executing the convolution, and carrying out the channel compression. The method comprises the steps that feature maps of all scales are enhanced, the enhanced feature maps of all the scales are input into a Neck network, cross-scale fusion is conducted through a USFF module and a C2FGCSA module, fusion features are input into a decoupling Anchor-Free detection head, a WIoU regression loss function is used for training a model, and the trained model is used for detecting the container lockhole. According to the invention, the container lockhole identification and positioning precision is improved.

The invention belongs to the technical field of bamboo wood processing, and provides a method for processing large-diameter bamboo wood slicing waste into paper pulp and a production system. According to the method, slicing waste generated during slicing of the large-diameter bamboo wood is used as a raw material for pulping treatment, paper pulp is obtained, and during slicing, the included angle between the tangent plane of a plane or a curved surface formed by the track of relative movement between a slicing knife edge and the bamboo wood and internode bamboo fibers is 0-1 degree. When the method and the production system provided by the invention are used for slicing the large-diameter bamboos, sliced products (bamboo strips or bamboo chips) meeting specification requirements are obtained; according to the method, slicing waste (such as bamboo shavings, bamboo core barrels and bamboo fragments) generated in the bamboo slicing process can serve as high-value raw materials to be used for preparing paper pulp with the long fiber length (the average fiber length reaches 2 mm), the defect of fiber raw materials in the papermaking industry is effectively overcome, and meanwhile the utilization rate and economic benefits of large-diameter bamboo raw materials are effectively increased.

The application discloses a kind of data augmentation methods for multi-attribute word extraction, it is related to data processing field, including: first, for the original training sample sentence containing multiple attribute words, attribute word labeling is carried out based on text fragment;Then, according to the labeling result, construct multi-attribute word label data set;Finally, each subset in the multi-attribute word label data set is combined with the original sample sentence to generate new training samples;The application reduces the cost of artificial data labeling, alleviates the problem of insufficient labeled data, designs a data augmentation method based on text fragment label in the multi-attribute extraction application scenario, synthesizes new training samples, and expands the training data;The augmented training data set enables the model to learn more features from the data, enhances the robustness of the model, prevents overfitting, and improves the generalization ability.

The application belongs to the field of image processing, and discloses a medical image pathology detection model training method based on few-shot learning, which comprises the following steps: S10, adopting a meta-learning paradigm, randomly sampling and constructing a few-shot training segment from a training data set containing multiple class-labeled images, each segment simulating a real few-shot classification scene and containing an example set and a target set; S20, constructing a shared feature encoder to process the example set and the target set in each segment; S30, generating a structured class anchor point based on a graph neural network for each class in the example set of the current segment; performing adaptive allocation of sample weights of the example set based on a meta-attention mechanism; S40, introducing a dynamic memory bank and an uncertainty perception mechanism to quantize the uncertainty of the class anchor point and integrate historical knowledge; S50, calculating a classification loss and updating the model to obtain a medical image pathology detection model. The application realizes accurate identification of pathological features of medical images under few-shot conditions.