Patents
Literature
Patsnap Copilot is an intelligent assistant for R&D personnel, combined with Patent DNA, to facilitate innovative research.
Patsnap Copilot

16911results about How to "Reduce occupancy" patented technology

A hidden flip ladder and oil field operation equipment

The invention relates to a hidden turnover ladder. The hidden turnover ladder is characterized by comprising an upper ladder assembly and a lower ladder assembly, wherein the upper ladder assembly comprises a baseplate, and a pair of mounting vertical plates symmetrically arranged on the baseplate; multiple stages of upper antiskid pedals are arranged between the mounting vertical plate; first pin shaft holes are formed in the lower front parts of the mounting vertical plates; the lower ladder assembly comprises a pair of step type mounting side plates; connecting lug plates are arranged at the upper ends of the mounting side plates, and are rotationally connected with the first pin shaft holes through first pin shafts; and multiple stages of lower antiskid pedals are arranged between the mounting side plates. The hidden turnover ladder has the beneficial effects of less occupied space, convenience for use and folding capacity in no use. The invention further discloses oil field operation equipment adopting the turnover ladder; and the oil field operation equipment has the following beneficial effects: second pin shafts are arranged for connecting a car body with the hidden turnover ladder, so that the turnover ladder can be accommodated in a turnover ladder accommodating cavity after two times of overturning, and the effect of convenience for use and transportation is achieved.
Owner:YANTAI JEREH PETROLEUM EQUIP & TECH CO LTD

Method and apparatus for quantizing and compressing neural network with adjustable quantization bit width

The invention relates to the technical field of neural networks, and specifically provides a method and apparatus for quantifying and compressing a convolutional neural network. The invention aims to solve the existing problem of large loss of network performance caused by an existing method for quantifying and compressing a neural network. The method of the invention comprises the steps of obtaining a weight tensor and an input eigen tensor of an original convolutional neural network; performing fixed-point quantization on the weight tensor and the input eigen tensor based on a preset quantization bit width; and replacing the original weight tensor and the input eigen tensor with the obtained weight fixed-point representation tensor and the input feature fixed-point representation tensor to obtain a new convolutional neural network after quantization and compression of the original convolutional neural network. The method of the invention can flexibly adjust the bit width according to different task requirements and can realize quantization and compression of the convolutional neural network without adjusting the algorithm structure and the network structure so as to reduce the occupation of memory and storage resources. The invention further provides a storage apparatus and a processing apparatus, which have the above beneficial effects.
Owner:INST OF AUTOMATION CHINESE ACAD OF SCI

Neural network model compression method and device

The invention discloses a neural network model compression method and a neural network model compression device. The neural network model compression method comprises the steps of: determining a model parameter set of each neuron layer for each neuron layer in a neural network model, wherein the model parameter set includes a plurality of model parameters; carrying out first transformation on the plurality of model parameters to generate a plurality of intermediate parameters; quantizing the plurality of intermediate parameters according to a preset quantization step size to obtain a plurality of quantization parameters; selecting a plurality of sampling quantization points from the plurality of quantization parameters according to a preset quantization bit number; generating quantized values of the plurality of model parameters according to values of the plurality of quantization parameters and the plurality of sampling quantization points; and compressing and storing the plurality of model parameters according to the quantized values. The neural network model compression method can better maintain the effect of the model, greatly reduces the size of the neural network model, and reduces the occupation of computing resources, especially memory resources.
Owner:BEIJING BAIDU NETCOM SCI & TECH CO LTD

Method and device for clearing memory of mobile terminal and mobile terminal

The invention provides a method and a device for clearing memory of a mobile terminal and the mobile terminal. The method for clearing the memory of the mobile terminal includes steps: obtaining the memory size of the mobile terminal, and obtaining corresponding clearing threshold value according to the memory size of the mobile terminal; obtaining memory information occupied by application programs which survive backstage, judging whether the application programs which survive backstage reach the clearing threshold value according to the memory information, and if yes, confirming the application programs to be cleared, and killing processes corresponding to the application programs to be cleared so as to release memory space occupied by the application programs to be cleared. The method and the device for clearing the memory of the mobile terminal and the mobile terminal automatically release the memory space occupied by the application programs to be cleared, avoid operation complexity generated when a user manually clears the memory, can timely reduce the size of the memory of the mobile terminal, occupied by the application programs which survive backstage, and improves running efficiency and performance of the mobile terminal.
Owner:BEIJING KINGSOFT INTERNET SECURITY SOFTWARE CO LTD

Sublevel open-stop and delayed filling mining method

InactiveCN102562065AImprove drilling positioning accuracyNo lossUnderground miningSurface miningTailings damFilling materials
The invention relates to a sublevel open-stop and delayed filling mining method. According to the method, cutting crosscuts are wholly undercut along the width of chambers and pillars so that the space of scope rock drilling is shared with ore drawing, drilling operations of vertical parallel blast holes are achieved in undercutting space, positioning accuracy of drilling the vertical parallel blast holes is high, the intervals among the blast holes are even, and the defect of high blasting boulder yield caused by drilling sector blast holes in traditional rock drilling cross cuts is avoided; during chamber (or pillar) scoping, driving of ore drawing gateways and ore drawing admission passages in fill materials of adjacent pillars (or chambers) isn't needed, and accordingly the safety is improved; simultaneously, ores are directly drawn from the undercutting space of the chambers and the pillars, the space of ore drawing is large, mutual interference is small, ore drawing efficiency is high, and ores are drawn completely without any dead space and secondary ore loss; artificial sill pillars are constructed, ore sill plates aren't needed to be reserved, the ores are free of primary loss, and the rate of resource recycling is high. Barren rocks and ore beneficiation tailings are used for filling gobs, the barren rocks are not taken out of pits, constructions of tailing dams and barren rock yards are reduced, and land occupation is small.
Owner:UNIV OF SCI & TECH BEIJING

Harmless disposal method of kitchen garbage

The invention relates to a harmless disposal method of kitchen garbage, which comprises the following steps of: (1) pretreatment: after unloading the kitchen garbage, performing solid-liquid separation, rough crushing, crushing and sorting and performing magnetic separation, and recycling metal substances mixed in the kitchen garbage; (2) hydrothermal hydrolysis: after the pretreatment, adopting the hydrothermal hydrolysis to fully inactivate viruses and bacteria in the kitchen garbage, separating out animal and vegetable fatty oil from the kitchen garbage, and hydrolyzing organic materials at the same time; (3) using the fatty oil obtained in the step (2) as an industrial fatty oil material or further performing deep processing to obtain fatty acid methyl ester or biodiesel; mixing and proportioning the pasty organic materials and then conveying into an anaerobic fermentation system; and conveying large granular organic materials into an aerobic biological fermentation device. The method provided by the invention realizes harmless disposal of the kitchen garbage mainly by the pretreatment and the hydrothermal hydrolysis process, increases the conversion rate of biological energy of the kitchen garbage and increases the economic benefit of kitchen garbage recycling products.
Owner:CHINA URBAN CONSTR DESIGN & RES INST CO LTD +2

Automatic debugging method and system of optical module

The invention provides a light module automation debugging system, wherein, the invention includes: a light module, which includes a digital controller used for reporting a performance amount or an alarm amount and revising value of the digital controller corresponding to the address of key parameter of the light module r according to the digital controller instruction delivered; a main control computer, which is used for downloading a reference value of the light module and receiving an actual value reported by the light module, and then comparing the reference value to the actual value, and if the values are not equal, finds out the digital controller requiring adjustment corresponding to the address of the key parameter of the light module and delivering a numerical instruction of the revised digital controller; and a debugging motherboard, which is used for bearing the light module and is connected with the main control computer through a interface for realizing a communication between the light module and the main control computer. Furthermore, the invention also provides a method for carrying out the light module automatic debugging, thus a production efficiency of the light module debugging is improved, and artificial quality hidden dangers in the process of the debugging are reduced.
Owner:ZTE CORP

Distributed system multilevel fault tolerance method under cloud environment

The invention provides a distributed system multilevel fault tolerance method under a cloud environment, which comprises a distributed application collaboration algorithm based on a virtual machine disk snapshot, which can back up the I/O state and the dependent operating system environment; a hierarchical fault detection and recovery mechanism, which can detect a physical layer, a virtual layer, a cloud platform layer, a virtual machine OS layer and an application layer fault in real time, and adopt the matched fault recovery method for different faults. Thus, the fault detection and recovery can be refined to modules, and the strategy of top-down stepwise recovery is adopted to minimize the recovery overhead; based on the virtual fault tolerance cluster service deployment strategy of the template, a user can use the virtual machine template to perform one-click deployment on the virtual machine fault tolerance cluster and upload the operation to perform collocation, and use the authorized fault-tolerant PaaS service. The invention can effectively solve the problems that the existing cluster deployment is complicated and the fault tolerance overhead is expensive, and can cope with the distributed application fault at all levels under the cloud computing environment in all directions.
Owner:HUAZHONG UNIV OF SCI & TECH

Data disaster tolerance system suitable for WAN

InactiveCN101316274ATo achieve the purpose of disaster recoveryReduce data volumeTransmissionRedundant operation error correctionComputer moduleGeolocation
The invention discloses a data disaster tolerant system which is applicable to a wide area network, pertains to the technical filed of computer information storage and solves the problem that the current data disaster tolerant system does not take the effect of data transmission links and the need of lots of redundant data storage space in the wide area network into consideration. The system of the invention comprises M local storage clients connected to the wide area network and N storage servers in different geographical positions; every local storage client consists of a web manager, a snap module, a mapping file redundant array module, a compression module, a decompression module, an encryption module, a decryption module, a backup distribution module and a recovery and integration module; the storage server consists of a failure detection module, a distributed arbitration module, a recovery module, a distributed copy service module, a backup and recovery service module. The system of the invention combines disaster tolerant coding and copying, reduces the amount of redundant data and effectively utilizes storage space. Meanwhile, data to be transmitted is compressed and encrypted, thus improving the safety and reliability of data on network transmission, effectively utilizing network bandwidth and reaching excellent disaster tolerant effect.
Owner:HUAZHONG UNIV OF SCI & TECH

Method and system for dynamic migration of WAN virtual machines

The invention provides a method for dynamic migration of WAN (Wide Area Network) virtual machines. The method aims to perform dynamic migration of one or more virtual machines from a source host to a destination host in a wide area network. The method comprises the following steps of: establishing a distributed index of the internal memory and external memory data of all virtual machines in operation in a data center to which the destination host belongs, thus forming a destination host environment; in the migration process of the virtual machines, retrieving the internal memory and external memory data of the virtual machines in the source host in the destination host environment, and transmitting the internal memory and external memory data not present in the destination host environment only through the wide area network; and simultaneously, in the transmission process, marking internal memory and external memory dirty blocks by the virtual machine manager of the source host, iteratively sending the internal memory and external memory dirty blocks until reaching a stop condition, and then stopping the transmission of the internal memory and external memory, causing the virtual machines in the source host to enter a stop stage, and starting the virtual machines by the destination host. The method is capable of reducing the network bandwidth occupied by the migration of the virtual machines in the wide area network and shortening the time of the dynamic migration of the virtual machines.
Owner:CHINA STANDARD SOFTWARE

Overhead power transmission line optimized line selection method based on airborne laser radar data

The invention discloses an optimal route selection method for overhead transmission line route. Onboard lidar equipment is adopted to acquire laser point cloud data and aerial digital photo data of the transmission line route corridor range; the onboard lidar data is processed after the wave filtering and the sorting of the laser point cloud data, and the points of the ground surface are made into a digital elevation model with high precision; then data processing is carried out by utilizing the data of the digital elevation model with high precision and ortho-rectification is carried out to the aerial photo by utilizing the internal and external orientation elements of the aerial digital photo to generate digital orthophoto maps; through the overlying of the digital elevation model and the digital orthophoto maps, the tridimensional visualization of the transmission line route corridor can be realized to optimize the transmission line route selection; finally prearrangement of power pole and power pole arrangement are carried out according to the data of plane cross sections. The route selection platform of the invention has simple operation and lifelike tridimensional scene, thus being convenient for full roaming and multi-view observation and greatly improving the efficiency of the inner plane cross section survey operation. Compared with the optimal route selection technology based on the aerial photographing measuring method, the efficiency of the inner plane cross section survey operation can be improved by about 75 percent.
Owner:GUANGXI ELECTRIC POWER IND INVESTIGATION DESIGN & RES INST
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products