四段式内啮合液压串泵

音箱（BS-52）

四段式内啮合液压可改串泵

WATERPROOFING APPARATUS

The present invention provides a waterproofing apparatus for an electronic device. The waterproofing apparatus comprises a sensing unit disposed in a case of an electronic device and used for outputting a corresponding sensing signal upon sensing that water permeates into the electronic device; and a powering-off unit used for generating a powering-off signal according to the sensing signal to power off the electronic device. In the present invention, power supply of a circuit board of an electronic device is immediately and automatically cut off upon it is sensed that water permeates into the electronic device, so that the problem that the circuit board of the electronic device is likely shorted out after water comes into the electronic device is solved, thereby achieving the effect of avoiding losses to a great extent.

音箱（SV-01）

平面端盖（五段式内啮合液压泵）

五段式内啮合液压串泵

Data-driven software memory related fault detection method

The invention belongs to the technical field of source code fault detection, and discloses a data-driven software memory related fault detection method. A code text is processed and cleaned, a TDTBert model based on Bert is used for pre-training, multi-modal feature vectorization is carried out on code text features and code syntax tree structural features, an expression mode of software fault features on text features and syntax tree features is generated, then feature enhancement and fusion are carried out, and software fault features are generated. Vectorized software fault features and fault information serve as a data set to train a TDTBert model, a fully-trained convergence neural network model is obtained and applied to fault detection related to the software memory, and the vector representation and recognition of fault types related to the software memory are good.

Green tide monitoring and drift path tracking method and system based on fluctuation glider

The invention discloses a green tide monitoring and drift path tracking method and system based on a fluctuation glider. The method comprises the following steps: arranging a plurality of sets of green tide monitoring fluctuation gliders in a monitored sea area to form a net-shaped structure; according to the monitoring data of the marine ecological background field, machine learning of a green tide monitoring algorithm is carried out, and an operation mode of a fluctuation glider is set; according to the monitoring data of the marine ecological background field, judging whether a green tide breaks out or not, if the judgment result is that the green tide breaks out, executing green tide early warning, and according to the effective wave height, judging whether to track a green tide drift path or not, and if the green tide does not occur, continuing to execute the monitoring of the marine ecological background field. By utilizing the technical characteristic that the fluctuation glider ...

INTELLIGENT SLAG DRAGGING ROBOT FOR METAL MELTING

An intelligent slag dragging robot for metal melting, comprising a control cabinet (1) and a slag dragging mechanism. The control cabinet is used for controlling the slag dragging mechanism to achieve automatic slag dragging; the slag dragging mechanism comprises a slewing bearing (3) and a base (12); the base is fixed on the ground; the slewing bearing is mounted on the base; the slewing bearing is connected to a large arm pull rod support seat (11); the large arm pull rod support seat is connected to the bottom end of a large arm (5) and that of a pull rod (4); the top end of the large arm is connected to a cross beam (6); the cross beam is connected to a telescopic arm (7); a clamping jaw (8) is mounted at the lower end of the telescopic arm. The intelligent slag dragging robot achieves full-automatic slag dragging, avoids the potential safety hazards of personnel in slag dragging work, and also can reduce the loss of a metal stock solution.

Model pruning method for interpretable CNN (Convolutional Neural Network) classification model

The invention relates to a model pruning method for an interpretable CNN classification model, belongs to the field of image compression, and solves the problems that an existing deep CNN model is high in operation complexity, large in time and memory consumption and difficult to deploy on terminal equipment, and an existing model pruning algorithm is lack of interpretability. The method comprises the following steps: inputting a training picture into a neural network model to be pruned, and extracting a feature map matrix of each convolutional layer; carrying out up-sampling on the feature map matrix to the size of an input picture, and carrying out normalization operation to construct a saliency map; multiplying the saliency map by the input picture element by element, and constructing a weighted input picture; carrying out element-by-element subtraction on the input picture and the weighted picture, and constructing an attention region shielding picture; and inputting the attention occlusion ...

Monorail trailer capable of automatically changing rails

A monorail trailer capable of automatically changing rails comprises a carriage and a hanging wheel set located at the top of the carriage, the hanging wheel set can run on a special rail, the special rail comprises a main rail, an auxiliary rail and a guide rail, the hanging wheel set comprises a front hanging wheel set and a rear hanging wheel set, the front hanging wheel set and the rear hanging wheel set are in linkage through a hydraulic pipeline, and the front hanging wheel set is provided with main wheels, auxiliary wheels and idler wheels. The rear hanging wheel set is provided with a main wheel and auxiliary wheels, the main wheel is fixed and can roll on the main rail, the auxiliary wheels are arranged beside the main wheel side by side and fixed to the inner swing arm so that the auxiliary wheels can rotate along with the inner swing arm to change the height of the auxiliary wheels within a certain range, and the idler wheels are fixed to the outer swing arm and can rotate along ...

Double-rotation-direction internal gear pump

The internal gear pump comprises a pump body and a pump cover which are oppositely arranged, a working cavity is formed in the pump body, a gear, a gear ring and a radial clearance compensation structure are arranged in the working cavity, the gear is arranged on the gear, the gear ring is arranged outside the gear in a matched mode, one side of the gear is meshed with the gear ring, and the other side of the gear is meshed with the pump cover. The radial clearance compensation structure is arranged in a clearance between the other side of the gear ring and the gear ring; and the oil inlet and the oil outlet are an axial oil inlet and an axial oil outlet which are arranged on the pump cover. Due to the adoption of axial oil inlet and outlet design, radial leakage points of the gear ring can be reduced, the volume efficiency of the oil pump can be improved, the machining and manufacturing cost of radial holes of the gear ring can be reduced, the double-rotation internal gear pump is matched ...

Resource allocation method based on user requirements under non-orthogonal multiple access

The invention belongs to the technical field of information processing and communication, and particularly relates to a resource allocation method based on user requirements under non-orthogonal multiple access, which utilizes reinforcement learning to solve the basic idea of NOMA (Non-Orthogonal Multiple Access) resource allocation: obtaining a resource allocation optimization function by taking spectrum utilization rate or energy efficiency as an optimization target and taking user service quality as a constraint; then defining a multi-agent user state space, an award and an action space, obtaining state space information through relatively small communication overhead, obtaining one-dimensional state space data, and obtaining a frequency spectrum and power distribution strategy according to the state space information; and finally, finding an optimal resource allocation strategy by training the deep neural network. According to the method, service classification is completed through ...

Differential privacy anonymization method based on attribute correlation

The invention belongs to the technical field of information security, and discloses a differential privacy anonymization method based on attribute correlation. The method comprises the following steps: firstly, analyzing the correlation among attributes through data distribution of a data set, calculating the number of each equivalent group of each attribute group by utilizing permutation and combination to find out all minimum attribute combinations, if the number of each equivalent group of the attribute combinations is greater than or equal to 30, and when any attribute is added, the equivalent groups with the number less than 30 exist in each equivalent group, and if the number of each equivalent group in each attribute combination is greater than or equal to 30, the minimum attribute combinations are found out; and the attribute combination is a minimum equivalence group. Then, obtaining the set condition of each attribute by intersecting all the minimum attribute groups, calculating ...

Battery replacement cabinet control method and system

The embodiment of the invention provides a battery changing cabinet control method, a battery changing cabinet comprises one or more battery grids, the battery grids are used for storing batteries and charging the batteries, and the method comprises the following steps: acquiring a first identifier corresponding to the battery changing cabinet specified by a user; based on the first identifier, determining a cluster in which the battery replacement cabinet specified by the user is located; the cluster comprises at least two battery changing cabinets; determining at least one empty battery grid from the battery changing cabinet in the cluster, and controlling a cabinet door of the empty battery grid to be opened; and determining at least one battery cell with a preset electric quantity battery from the battery replacement cabinet in the cluster, and controlling the opening of a cabinet door of the battery cell.

Ploughing device with crushing function

The invention relates to the technical field of plowing devices, in particular to a plowing device with a crushing function, which comprises a plow frame, a plurality of plow shares are arranged on the plow frame in an array mode, a first connecting support and a second connecting support are respectively arranged at two ends of the other side of the plow frame, and a collecting structure is arranged between the first connecting support and the second connecting support. The collecting structure comprises a material collecting structure, the material collecting structure comprises a first rectangular frame, connecting vertical plates are arranged at the two ends of the first rectangular frame, a plurality of first transverse rods are arranged in the first rectangular frame in the length direction of the first rectangular frame in an array mode, and a flapping structure is arranged at the upper end of the first rectangular frame and comprises a second rectangular frame. A plurality of first ...

Sandstone pile arrangement structure and method for dissipating excess pore water pressure of soil body around tunnel

The invention relates to a gravel pile arrangement structure and method for dissipating excess pore water pressure of soil around a tunnel, the gravel pile arrangement structure comprises gravel pile arrays which are distributed on the two sides of the tunnel in a rectangular mode, and each side comprises m columns of gravel mixed piles; in the extension direction of the tunnel, the distance S longitudinal of the adjacent gravel mixed piles is determined according to the following formula: in the formula, d is the pile diameter of the gravel mixed piles, and n is the borehole diameter ratio and is determined according to the following formula: in the formula, t is the time from the arrival of a shield tunneling machine to the separation of a shield tail, U is the consolidation degree of the stratum around the gravel mixed piles, and ch is the transverse consolidation coefficient mgt; when 1 is greater than 1, the distance S < transverse > 'between the adjacent gravel mixed piles which are ...

Network attack detection method based on deep reinforcement learning

The invention discloses a network attack detection method based on deep reinforcement learning. The method comprises the following steps: preprocessing an original data set, and constructing an Agent, including initializing an environment in which the Agent is located, and specifying an interaction mode, a training strategy and a value function of an intelligent agent and the environment; and selecting the features according to the state, and inputting the selected features into a detection model for prediction. And the detection result is used as feedback to be transmitted back to the agent training module, Q (s, a) of the action is calculated, and a Q table is refreshed. The steps are repeated until the number of features contained in the optimal feature subset reaches the maximum, namely, the model is converged; or the training step length is completed, and an optimal feature subset is generated. The processing method designed for the novel features can reflect the importance of the ...

Cerebral arterial thrombosis prediction method based on ensemble learning

The invention belongs to the field of data mining, and discloses an ischemic stroke prediction method based on ensemble learning. According to the method, firstly, missing value filling and integer coding of category data are carried out on EHR data, then the data are copied into k copies, and k base classifiers are correspondingly selected as a first layer of an integrated model; and carrying out feature selection based on a genetic algorithm on each part of data, and using a result to train a corresponding base classifier. After the process is completed, prediction results of the k base classifiers are regarded as k evaluation indexes to serve as input data of the second layer of the integrated model, namely an integrated classifier, and the data are used for training the integrated classifier. And finally, a two-layer integrated model is obtained and is used for predicting the risk of ischemic cerebral apoplexy.

Federal learning hybrid aggregation mechanism under block chain architecture

The invention belongs to the technical field of machine learning privacy and security, and provides a federal learning hybrid aggregation mechanism under a block chain architecture. Local servers are used as management and miner nodes, and a network is divided into a plurality of local regions composed of the local servers and terminal devices and an inter-domain network formed by connecting the local servers. In the local region, gradient selection is carried out by using a coding algorithm meeting local differential privacy, and gradient index attacks are defended; gradient perturbation is carried out by using a perturbation algorithm satisfying local differential privacy, and gradient content attack is defended; encrypting the gradient by using a public key of a local server to defend eavesdropping attack; and shuffling and sampling the user update by using the shuffling server to defend the user identity attack. The local server is used for decrypting and aggregating intra-domain updates ...