48results about How to "Break through performance bottlenecks" patented technology

Disclosed is a parallelization method of distributed hydrological simulation under cluster environment. Confluence process calculation which is a limiting factor in parallel computation of the distributed hydrological simulation and slope process calculation with good parallelism are distinctively treated, a parallel algorithm under the cluster environment is researched and developed, so that calculation efficiency is improved. The technical scheme includes that the method comprises the steps: carrying out substream partition and classification; serving the substream as a template, subdividing input data and storing the data into a data bank; serving the area of the substream as a measurement index of calculated amount, at the same time, considering a topological relation among substreams and carrying out partition of calculation tasks; serving the substream as a unit, carrying out parallel computation under the cluster environment, wherein the slope process calculation adopts a static scheduling, and the watercourse process calculation adopts a dynamic scheduling; and writing a result into the data bank after the calculation is finished. According to the parallelization method of the distributed hydrological simulation under the cluster environment, performance bottlenecks of an existing parallelization method of the distributed hydrological simulation under the cluster environment are broken through, parallel performance is remarkably improved, and application prospect in aspects of large-watershed high-resolution distributed hydrological simulation is wide.

The invention relates to a self-enhanced polylactic acid fiber and a preparation method thereof. The self-enhanced polylactic acid fiber provided by the invention is a uniform mixture of a vertical compound crystal polylactic acid micro-fiber and a polylactic acid substrate. In 100 weight parts of self-enhanced polylactic acid fiber, 1-30 parts of vertical compound crystal polylactic acid micro-fiber and 70-99 parts of polylactic acid substrate are contained. The preparation method of the fiber comprises the following steps: firstly, respectively carrying out vacuum drying on PLLA (poly L lactic acid) and PDLA (poly D lactic acid); physically mixing PLLA with PDLA after being dried; fusing the mixture; collecting primary fibers under the conditions that a spinning temperature is 225-245 DEG C and a spinning speed is 500-2500 meters per minute; and thermally drafting and thermally shaping the primary fibers. Compared with the common polylactic acid fiber, the self-enhanced polylactic acid fiber disclosed by the invention has the advantages of higher constant temperature mechanical strength, higher high temperature mechanical strength and lower boiling water shrinkage, so that the use demand is met and the application field is expanded.

The invention relates to the technical field of metal passivators, and particularly relates to a chromium-free composite passivator and a preparation method thereof. The chromium-free composite passivator provided by the invention is prepared from the following components in percentage by mass: 3.0-10% of gamma-methylacryloyloxytrimethoxy silane, 0.1-3.0% of sodium molybdate, 0.1-5.0% of cerium chloride, 0.5-4.0% of phosphoric acid, 0.5-4.0% of acetic acid, 15-20% of water-soluble epoxy resin, 5-15% of water-soluble acrylic resin, 10-30% of water and 20-40% of ethanol. The chromium-free composite passivator is suitable for treating steel, zinc and zinc alloy, or aluminum and aluminum alloy. The chromium-free composite passivator contains the water-soluble high-molecular resins and inorganic corrosion inhibitor, and is mixed with ethanol to form a colorless, innoxious, environment-friendly, transparent and clear solution. At room temperature, the chromium-free composite passivator is applied on the surface of steel, hot-galvanized zinc and zinc alloy, or aluminum and aluminum alloy to generate chemical crosslinking reaction, and is cured to form a film.

The invention relates to an application service system for monitoring a power charging-exchanging station and an interprocess communication method. The system comprises a service request end, a service supply end and an application service bus, wherein the service request end is used for putting forward a service request from the application service bus; the service supply end is used for supplying service corresponding to the service request sent from the application service bus; the application service bus is used for managing the application service, interaction and processing of information, and monitoring the operation conditions of each node and conducting corresponding control; and the service request end and the service supply end are respectively in communication connection with the application service bus through a service bus adaptor and a bus standard interface. The application service system realizes plug and play to various components conforming to soft bus interface standards and is easy for integration and customization of a large-scale software system; the standardization of component interfaces gives rise to hardware or industrialization of production, therefore the production efficiency is increased; and redundancy deployment of double application service buses is adopted, the performance bottleneck of single bus load is broken through, single-point failures are overcome and the applicability of the system is enhanced.

The invention relates to a breast tumor classification algorithm based on a convolutional neural network VGG16. The algorithm comprises the following steps that: data preprocessing: for a dataset which presents a data imbalance state, carrying out imbalance processing and data enhancement processing; the establishment of the convolutional neural network: 1) network pre-training: utilizing the VGG16 to carry out network training on an ImageNet large natural image dataset, and storing trained weight; 2) network key node selection: utilizing different layers of the VGG16 network to carry out feature extraction on a breast tumor DDSM (Digital Database for Screening Mammography) dataset, applying the same SVM (Support Vector Machine) classifier for classification for extracted features, and selecting a layer with highest classification performance as a node constructed by a new network; and 3) connecting two layers of full connection and one layer of softmax to form a new network behind thenode constructed by the selected network; and carrying out migration learning.

The invention discloses a low-power-consumption 12-bit assembly line successive approximation analog-digital converter which comprises a 6-bit sublevel successive approximation analog-digital converter, an operational amplifier, a 7-bit sublevel successive approximation analog-digital converter and a digital delay calibration unit. The residual signal output end of the 6-bit sublevel successive approximation analog-digital converter is connected with the input end of the operational amplifier, digital signals of the 6-bit sublevel successive approximation analog-digital converter are output to the high 6-bit input end of the digital delay calibration unit, the output end of the operational amplifier is connected with the input end of the 7-bit sublevel successive approximation analog-digital converter, and the digital signal output end of the 7-bit sublevel successive approximation analog-digital converter is connected with the low 6-bit input end of the digital delay calibration unit. The converter effectively achieves combination of assembly line analog-digital converters and successive approximation comparators, effectively improves comprehensive performance of the analog-digital converters, is low in power consumption compared with like analog-digital converters and has wide application prospects in the deep submicron process.

The present invention provides a hot-press sintering device. The device comprises a top electrode, a bottom electrode, a water-cooling vacuum chamber and a mould group. The mould group comprises a mould main body, an upper pressing head and a lower pressing head, the mould main body is provided with a through hole in a height direction, the sum of the heights of the upper pressing head and the lower pressing head is smaller than the height of the mould main body, when the hot-press sintering device is operated, the mould set is arranged in the water-cooling vacuum chamber, the top electrode and the bottom electrode press the upper pressing head and the lower pressing head into the through hole of the mould main body until the upper pressing head and the lower pressing head to be leveled with the upper end face and the lower end face of the mould main body, and the through hole of the mould main body is internally provided with a sample chamber holding samples. The present invention further relates to a method for preparing thermoelectric materials through adoption of the device mentioned above, and thermoelectric materials of a multi-scale micro-nano porous structure with a chemical formula of Mg3.2-xMnxSb1.5-yBi0.5Tey, wherein 0.0125<=x<=0.1, and 0.01<=y<=0.05.

The invention discloses a blade, a modeling method and manufacturing method thereof and an air compressor. The modeling method comprises the steps that a flowing channel is sequentially divided into an inlet semi-vaneless area, a channel expansion area and an outlet semi-vaneless area in the gas flowing direction; front edge side end points of a suction surface molded line and a pressure surface molded line of the flowing channel are determined; blade molded lines forming the boundaries of the inlet semi-vaneless area, the channel expansion area and the outlet semi-vaneless area are sequentially determined, and the channel expansion angle of the whole flowing channel is regulated and controlled; on the basis of the determined molded lines, modeling of the primitive-level blade profiles is completed, and then the primitive-level blade profiles are stacked to form the three-dimensional blade. The modeling method has the advantage that the diffusion degree of an expansion channel is controllable, the quality of flowing airflow in the expansion channel of the axial flow air compressor blade is improved helpfully, pneumatic matching between blade rows is improved, and the efficiency of the axial flow air compressor is improved; and the modeling method has good applicability to radial/axial diffuser blades of a centrifugal air compressor and high-low pressure stage rotor/stator blades of a turbine.

The invention provides an air-sea wireless channel estimation method based on deep learning, and the method comprises the steps: generating channel transmission function (CTF) data by building an offshore wireless channel simulation platform; collecting an orthogonal frequency division multiplexing pilot frequency receiving signal, and calculating least square estimation of the CTF at a pilot frequency position; preprocessing the least square estimation of the CTF data and the CTF at the pilot frequency position to construct a training data set; constructing an efficient channel estimation convolutional neural network, performing offline training on the efficient channel estimation convolutional neural network according to the training data set, and updating weight parameters of the efficient channel estimation convolutional neural network; and solidifying weight parameters of the efficient channel estimation convolutional neural network, and performing online testing. According to the air-sea wireless channel estimation method provided by the invention, sparse features on a channel delay domain are extracted by using trainable sparse transformation, so that model parameter redundancy can be reduced; meanwhile, according to the method, interpolation calculation of any up-sampling factor is achieved through a sub-pixel convolution layer and a cutting layer, the model calculation complexity can be reduced, and the mean square error performance bottleneck is broken through.

The invention relates to a pressure regulating structure of a variable loading type pressure regulator. The pressure regulating structure is arranged in a gas pressure regulator, and a pressure regulator diaphragm is arranged in the gas pressure regulator. The pressure regulating structure comprises a differential pressure elastic air bag arranged on the pressure regulator diaphragm and a pilot valve connected to the differential pressure elastic air bag through an air guide signal pipe, wherein the differential pressure elastic air bag is a circle of completely sealed closed circular ring, and an elastic air bag cavity D is arranged in the differential pressure elastic air bag; and a pilot valve air inlet A is arranged in the pilot valve, the pilot valve comprises a pilot valve body, a piston is arranged in the pilot valve body, and the air guide signal pipe is connected into the pilot valve body and is controlled to be on and off through the position of the piston. The self-operatedvariable loading structure does not need external force, has a self-adaptive adjusting function, is safe, reliable and good in control performance through the loading mode of the differential pressureair bag and the control mode of the differential pressure pilot valve and the air bag, provides the most suitable loading area and is high in regulation precision.

The invention relates to a good-touch glass fiber reinforced polypropylene compound and a preparation method thereof. The good-touch glass fiber reinforced polypropylene compound comprises polypropylene, a toughening agent A, a toughening agent B, glass fibers, a maleic anhydride graft, a nucleating agent, an antioxidant and a light stabilizer. According to the invention, by adding the uniquely compounded flexibilizer, the dissociation viscosity of a produced polypropylene material with good touch feeling is obviously improved, the requirements of the public material standard PV1306 can be met, the bottleneck of performance is broken through, and the application field of reinforced polypropylene in plastic parts on automobiles is greatly widened.

The invention discloses a middleware method for realizing searchable encryption relational database caching, which specifically comprises the following steps of: 1, establishing mapping between a plaintext database and a ciphertext data connecting middleware to the plaintext database, and integrally converting the middleware and the ciphertext database into the searchable ciphertext database;using a unified searchable encryption algorithm to add, delete, modify and check newly stored data to realize fusion of a new database and an old database; Step 2, establishing a user query cache byutilizing an NOSQL data carrying out SQL analysis and routing on the middleware, cache hit and miss on the middleware, calculating column access frequency, and carrying out cache monitoring; Andstep 3, for each query operation of the user, judging whether the query request hits the cache or not, if so, returning a result from the memory database, otherwise, connecting the ciphertext database, and returning the query result. According to the method, the query speed of the searchable ciphertext database is increased and the concurrency problem is relieved while the security is ensured.

A real-time detection tracking method for a moving object of a space-borne satellite image belongs to the technical field of the satellite image and in particular relates to a detection tracking method for the moving object of the space-borne satellite image. The invention can solve the problems of error tracking caused by a lot of moving noisy points brought by image stabilization, tracking lossrate during multi-target tracking and performance of a detection tracking algorithm. The algorithm breaks through the performance bottleneck, and is capable of realizing real-time monitoring of a space-borne satellite and reducing the labor cost; an abnormal condition can be warned in time through a precise detection tracking result, man power is not needed for monitoring all the time, and the method can be applied to the military field for detection tracking of the moving object such as an aircraft carrier and a warship.

The invention provides a network message matching method of a custom protocol application layer based on K-nearest neighbor, and belongs to the field of data transmission. The method comprises the following steps: constructing a protocol type storage structure according to historical network messages, carrying out sequential numbering according to timestamps, obtaining a structure feature value and a value array of each protocol type, obtaining prediction protocols of K neighbors based on a K neighbor model, inputting the prediction protocols and feature values into a linear regression equation, and carrying out linear regression on the prediction protocols of K neighbors according to the linear regression equation; obtaining similar weights of a to-be-matched protocol and K prediction protocols, sorting the similar weights, obtaining a source IP and a destination IP of a to-be-matched message, querying corresponding protocol types, and forming U with the prediction protocols; matching the network message to be matched with the protocol in the U one by one, and when the matching is successful, transmitting data and returning a protocol type; and when the matching fails, the local protocols except the U are polled and matched until the matching is correct. According to the invention, the data transmission efficiency is improved while the transmission reliability and security are ensured.

The invention discloses a rapid light-weight face recognition method, and the method comprises a face detection step: carrying out the detection of a to-be-detected video, obtaining a detection target, and obtaining a face image of the detection target; a key point extraction step: performing human face key point extraction on the human face image, and performing human face alignment based on thehuman face key points; a living body verification step: performing living body verification on a corresponding detection target based on the aligned face image; a feature value calculation step: calculating a face feature value of the face image according to the aligned face image; and a face recognition step: performing face recognition on the detection target based on the face feature value, wherein the face detection step, the key point extraction step, the living body detection step, the feature value calculation step and the face recognition step are executed asynchronously, and the living body verification step and the characteristic value calculation step are executed in parallel. According to the method and device, the living body verification step and the feature extraction step are executed in parallel, so that half of time is saved, and quick response is achieved during face recognition.

The invention provides a distributed scheduling method based on the same interface. The distributed scheduling method involves a scheduling engine, a service interface plug-in and a plurality of service implementation plug-ins and comprises the following steps that access requests are classified; the number of servers required by page views and the number of service implementation plug-ins required by the page views are calculated out; the servers of the server implementation plug-ins implementing the interface in a cluster are found; certain servers meeting requirements and having the lightest loads are found; the scheduling engine distributes the requests to the service implementation plug-ins of the certain servers, so that distributed computation and parallel computation are conducted, and results are gathered and output; after the requests are executed, the scheduling engine releases the service implementation plug-ins. According to the distributed scheduling method based on the same interface, dynamic increasing and decreasing of the service implementation plug-ins are achieved, performance bottlenecks of a single service plug-in are overcome, and therefore performance of an application system is promoted.