135results about How to "Optimize network structure" patented technology

The invention discloses a multi-stage target detection method and model based on CNN multistage characteristic fusion; the method mainly comprises the following steps: preparing a related image data set, and processing the data; building a base convolution nerve network (BaseNet) and Feature-fused Net model; training the network model built in the previous step so as to obtain a model of the corresponding weight parameter; using a special data set to finely adjust the trained detection model; outputting a target detection model so as to make target classification and identification, and providing a detection target frame and the corresponding precision. In addition, the invention also provides a multi-class target detection structure model based on CNN multistage characteristic fusion, thus improving the whole detection accuracy, optimizing model parameter values, and providing a more reasonable model structure.

The present invention discloses a feature extraction and state recognition method for one-dimensional physiological signal based on depth learning. The method comprises: establishing a feature extraction and state recognition analysis model DBN of a on-dimensional physiological signal based on depth learning, wherein the DBN model adopts a "pre-training+fine-tuning" training process, and in a pre-training stage, a first RBM is trained firstly and then a well-trained node is used as an input of a second RBM, and then the second RBM is trained, and so forth; and after training of all RBMs is finished, using a BP algorithm to fin-tune a network, and finally inputting an eigenvector output by the DBN into a Softmax classifier, and determining a state of an individual that is incorporated into the one-dimensional physiological signal. The method provided by the present invention effectively solves the problem that in the conventional one-dimensional physiological signal classification process, feature inputs need to be selected manually so that classification precision is low; and through non-linear mapping of the deep confidence network, highly-separable features / feature combinations are automatically obtained for classification, and a better classification effect can be obtained by keeping optimizing the structure of the network.

The invention relates to a method for preparing an oil absorption material, in particular to a method for preparing a carbon nanofiber aerogel oil absorption material from bacterial cellulose. According to the method, bacterial cellulose aerogel is modified to obtain the carbon nanofiber aerogel oil absorption material. The method comprises the following steps: culturing the bacterial cellulose; preparing the bacterial cellulose aerogel; and preparing the carbon nanofiber aerogel oil absorption material. The method for preparing the carbon nanofiber aerogel from the bacterial cellulose is simple, novel in design, high in environment-friendly biocompatibility and high in absorption efficiency, and the oil absorption material can be recycled for many times. The hydrophobic carbon nanofiber aerogel prepared by the process can widely absorb multiple organic solvents and oil products, and is high in recyclability and selectivity; the absorption rate is up to 50-300 times the self-weight of the hydrophobic carbon nanofiber aerogel.

The invention relates to a modified starch-based wood adhesive as well as a preparation method and application thereof, which belong to the fields of an adhesive and a preparation method thereof. The preparation method comprises the following steps of: after carrying out acidolysis-gelatinization pretreatment on starch as a main raw material, carrying out graft copolymerization on the starch in a seed emulsion polymerization process by adopting vinyl acetate monomers, and adding nano silicon dioxide sol and other assistants after cooling to finally obtain a wood starch adhesive with excellent performance. The main technical indexes of the adhesive prepared with the preparation method are all better than those of the traditional starch-based wood adhesive, and the adhesive has the advantages of no toxicity or pollution, high bonding strength, good water resistance, milky white and glossy color, high drying speed, long storage time and the like and can be widely applied to bonding decorative wood and artificial boards.

The invention relates to a blending modified hollow polysulfone fiber film which comprises the following components in percentage by weight: 3-40 percent of polysulfone, 0.5-20 percent of porogen, 50-90 percent of organic solvent and 0.5-20 percent of inorganic nano particles, wherein the inorganic nano particles are one or the mixture of more than two of silicon oxide nano particles, titanium oxide nano particles, silicon carbide nano particles and gamma-phase aluminium oxide nano particles; the viscosity-average molecular weight of the polysulfone is 40,000-200,000; the organic solvent is one of N-methyl-2-pyrrolidone, dimethyl acetamide or dimethyl sulfoxide; and the porogen is one or a mixture of any two of polyvinylpyrrolidone, polyethylene glycol and polyvinyl alcohol. The invention improves the surface property, the hydrophilic property and the mechanical strength of the film through modifying the hollow polysulfone fiber film, has simple film preparation process, and can realize industrial mass production.

The invention relates to a network adjustment method and device, and belongs to the technical field of local area network communication. The method comprises the following steps: after receiving a network adjustment instruction, scanning each wireless access point surrounding a terminal; determining a target access point which satisfies a preset condition among the scanned wireless access points; and after the connection between the connected wireless access points is disconnected, accessing the target access point, and then the terminal can automatically switch the accessed wireless access points according to an actual network environment, so as to achieve an effect of optimizing a network structure of a family local area network.

The invention discloses a human posture recognition method based on depth learning, which mainly solves the problems of large calculation amount and low accuracy of the current posture recognition technology. Firstly, the kinect V2.0 depth sensor is used to collect the motion and posture characteristics of several human body samples; RGB data and bone data of human motion posture are stored; skeletal images obtained from bone data after image preprocessing are used as training set and test set; the training set is inputted into a Posture based on Convolution Neural Network (CNN), which is specially used in the field of human posture recognition. In CNN, the classification result is obtained after training and adjusting the network structure and network parameters. The motion and posture characteristics of different human body samples are input into the classification network as a test set, and the most probable motion is the recognition result. The invention uses the convolution neuralnetwork to improve the identification accuracy, reduce the identification time, has low operation cost, and is simple and convenient to be applied in places such as intelligent home, safety monitoring, motion analysis and the like.

The invention relates to a rapid fault detection and power recovery system based on a rural power network, and a concentrator detection method thereof. The rapid fault detection and power recovery system comprises an automatic metering system, a communication interface and a terminal and further comprises an alarm module; processed data comes from the automatic metering system and a marketing management system; the processed data is fed back to a low-voltage comprehensive application management system; the terminal comprises a distribution transform monitoring terminal, a concentrator and a metering unit group; the minimum detection voltage of a voltage detection module of the concentrator is 10V; both the distribution transform monitoring terminal and the concentrator are each provided with a wireless communication module communicated with the automatic metering system, and periodically perform data communication with the automatic metering system; the heartbeat cycle of the distribution transform monitoring terminal is 15 m; and the heartbeat cycle of the concentrator is 5 m. According to the rapid fault detection and power recovery system based on the rural power network, and the concentrator detection method thereof disclosed by the invention, the alarm precision is finely divided into four kinds by upgrading software in the existing concentrator, completing the interface of the automatic metering system and shortening the data interaction cycle of the concentrator and an ammeter; different maintainers can be dispatched easily; and thus, active and rapid first-aid repair is realized.

This invention discloses a method for preparating a kind of inflatable vermiculite / poly-potassium acrylate-acrylamide super water-retention property water -absorbent composite material, which is made by graft copolymerizing acrylic acid, acrylamide and inflatable vermiculite powder. In this invention, inflatable vermiculite powder takes a art of 50%, whose distilled water absorption could amount to 850 times. The absorbency of inflatable vermiculite / poly-potassium acrylate-acrylamide super water -absorbent composite material is a little lower than that of pure organic resin polymer(it can fulfill normal using requirements), while its manufacturing cost is apparently less than that of pure organic resin polymer. Also it has very good water-absorbency and water-retention property, so it has very extensive use in agriculture and forestry, plant planting, soil improvement, food hygiene, buiding, petroleum chemical industry, and so on.

A preparation method of a magnetic bacteria cellulose aerogel oil absorption material, and relates to a method for preparing an oil absorption material. The aim of the invention is to provide the preparation method of the magnetic bacteria cellulose aerogel oil absorption material. A bacterial cellulose aerogel is modified into a carbon nanometer fiber aerogel oil absorption material, and magnetic particles are added to the prepared aerogel, so that magnetic substances can be conveniently used for salvage and recovery after oil absorption and are beneficial to collecting. The method comprises the steps as follows: 1, cultivation of bacterial cellulose; 2, preparation of bacterial cellulose aerogel; 3, preparation of the carbon nano fiber aerogel absorption material; and 4, preparation of the magnetic carbon nanofiber aerogel oil absorption material. The invention prepares magnetic carbon nanofiber aerogel from bacterial cellulose, the preparation method is simple, novel in design and environment-friendly, and has good biocompatibility, high absorption efficiency and convenience for fishing, and can be reused for many times.

This invention discloses a rare earth series extraction intelligent control system, which comprises operation station and spot executive system and is characterized by the following: fixing model station, together with operation, connected to the OPC communication module, wherein, the above two and the PLC controller are connected to the spot executive system through industry Ethernet and Profibus-DP. The invention also provides one method to execute the control system, which is characterized by the following: the model station feedbacks the difference value between the optimization parameter by experiences with spot parameters as the basis of the operation station. The invention also discloses one multi-media monitoring system to real time display the spot situations in workshops.

A number automatic routing method, an updating method, a withdrawing method, a router and a device is provided wherein the number automatic routing method includes: converting to-be-distributed number into number routing information to add to dynamic routing protocol; generating a dynamic routing protocol message according to the number routing information; transmitting the dynamic routing protocol message through an Internet Protocol (IP) network and distributing the number through the dynamic routing protocol message and the IP network. The method supports automatic routing of numbers such as E.164 and E.214 and SP through expanding current Multi-Protocol Border Gateway Protocol (MP-BGP) that optimizes network structure and improves efficiency.

The invention relates to a starch-grafted acrylic acid / modified tourmaline super absorbent polymer and a preparation method thereof, which relates to the super absorbent polymer and the preparation method thereof. The method solves the problems of complex technique, high cost, poor saline tolerance of products, low jelly strength, low water-absorption rate and poor thermal stability in the prior method for compounding inorganic mineral and the starch-grafted super absorbent polymer. The starch-grafted acrylic acid / modified tourmaline super absorbent polymer is prepared by a vinyl monomer, starch, a crosslinking agent, an initiator, modified tourmaline powder and deionized water. The method comprises the following steps: firstly, compositions required are weighed and taken; and secondly, the starch is pasted, added with the vinyl monomer and the modified tourmaline powder, and then added with the initiator and the crosslinking agent, and the starch-grafted acrylic acid / modified tourmaline super absorbent polymer is prepared after reaction, drying and crushing. The method has a simple and convenient technique and low cost; the saline tolerance of products prepared is good; the jelly strength is high; the water-absorption rate is improved by 10 percent; and the thermal stability is good at a temperature between 200 and 300 DEG C.

The invention discloses a base station planning method and device, terminal equipment and a storage medium, and relates to the field of communication. The method comprises the following steps: determining the position of a base station planning subarea in a target planning area based on a position determination rule of the base station planning subarea; according to the position, determining a target grid located in the base station planning sub-region from grids obtained by dividing the target planning region; when the proportion of a weak coverage grid in the target grid is greater than a predetermined proportion threshold, selecting a target position in the base station planning sub-region as the position of a to-be-built base station, the weak coverage grid being specifically a grid ofwhich the reference signal receiving intensity of the mobile communication network is lower than a preset signal receiving intensity, and the weak coverage grid being a grid of which the reference signal receiving intensity is lower than a preset signal receiving intensity; and further, optimizing the to-be-built base station in the target planning area. Through the scheme provided by the invention, the position of the to-be-built base station can be accurately obtained.

The invention provides an activated form gel type lithium ion battery polyelectrolyte film, a preparation method and application thereof. The invention prepares P(AN-VAc) powder by using monomer acrylonitrile and vinyl acetate or prepares P(AN-MMA) power by using acrylonitrile and methyl methacrylate; and prepares a P(AN-VAc) or P(AN-MMA) polymeric matrix film on the basis; an activated form polymer film is obtained by drying after the P(AN-VAc) or P(AN-MMA) polymeric matrix film is immersed into polymethyl methacrylate gel to be activated, and an activated form gel type lithium ion battery polyelectrolyte is obtained by immersing the activated form polymer film into an electrolytic solution. Compared with the unactivated P(AN-VAc) polyelectrolyte film, the activated form P(AN-VAc) / PMMA polyelectrolyte film obtained by adopting the technology of the invention has better absorbency, room-temperature conductivity, electrochemical stability and the interface stability of electrolyte and lithium electrode; the performance of a lithium secondary battery prepared by using activated electrolyte is obviously better than the performance of the lithium secondary battery prepared by using the unactivated polyelectrolyte; the invention has simple technology, short reaction time, and convenient and easy condition for technologized production.

The invention discloses a firewall policy configuration method based on service access data, and belongs to the technical field of data analysis application. The method includes steps: acquiring, extracting and classifying service access data including source IPs, source ports, target IPS, target ports, applications and protocols in multiple network paths through a network monitoring bypass mode;locking the normal service access data in the service access data obtained by first acquisition in the acquisition step as access rule data, comparing the service access data obtained by each subsequent acquisition in the acquisition step with the previously acquired and locked normal service access data, and locking the subsequently acquired data as access rule data; and when an access rule is issued to a firewall, firstly inquiring the access rule data locked in the access rule locking step, generating a forbidding rule of a firewall, and then issuing the forbidding rule of the firewall to the firewall.

The invention discloses a power price prediction method based on an improved deep belief network, and the method comprises the steps: dividing a data set and determining the input of network data according to the characteristics of electricity price data and the influence factors of electricity price, and carrying out the data preprocessing of an adopted data set; for the preprocessed data set, calculating a network error by using a second-order reconstruction error, and determining the number of layers of the model RBM; optimizing the number of neuron nodes in the network by using a '3 + 2 'search algorithm combining a trisection method and a bisection method; using a BP neural network and an SVR support vector regression machine used as regression layers of a DBN network, and using the number of layers of an RBM and the number of optimized neuron nodes to construct a DBN-BP model with an optimized structure and the DBN-SVR model with an optimized structure; and predicting the real-time electricity price data. According to the invention, the DBN model with an optimized structure is established, and different combination improvements are carried out on the regression layer of the network, so that the prediction precision of the DBN is improved, and the application prospect is very good.

The invention relates to a polysiloxane- and polyphosphonic acid-doped chitosan high-temperature proton exchange film and a preparation method thereof. The polysiloxane- and polyphosphonic acid-doped chitosan high-temperature proton exchange film is the product obtained through the following steps: (1) preparing chitosan solution; (2) preparing organic polyphosphonic acid solution and polysiloxane solution; (3) adding the polysiloxane solution in the chitosan solution at room temperature, and adequately and uniformly stirring to obtain the mixed solution; (4) placing the mixed solution in an ice-water mixed bath, slowly dropwise adding the organic polyphosphonic acid solution, after that, continuously stirring to obtain a sol; and (5) pouring the sol in a polytetrafluoroethylene die disc, drying and then carrying out heat treatment, cooling and then peeling to obtain the polysiloxane- and polyphosphonic acid-doped chitosan high-temperature proton exchange film. The preparation method disclosed by the invention is simple in preparation process and moderate in reaction conditions; moreover, the prepared high-temperature proton exchange film still has high proton conductivity under a high-temperature and low-humidity condition and even a water-free condition.

The invention relates to a device fault diagnosis apparatus and system. The device fault diagnosis apparatus comprises a management storage unit, a BP fault diagnostic device construction unit, a particle dimension determination unit, a particle initialization unit, a PSO-BP fault diagnostic device construction unit, a first diagnosis result generation unit, a particle position updating unit, a preliminary diagnosis model construction unit, an optimal particle selection unit, a final diagnosis model construction unit, and a fault diagnosis result generation unit. The BP neural network model fault diagnosis method is optimized by using a particle swarm algorithm; and on the basis of the traditional BP neural network, the particle swarm optimization (PSO) algorithm is introduced to change the original error iterative calculation mode. The global search capability of the particle swarm optimization algorithm is combined with the local fast search capability of the BP neural network to avoid a phenomenon of trapping into local optimization by the result. Therefore, the PSO-based optimized BP neural network model is established and the network structure is optimized; and the weight andthreshold of the BP neural network are optimized by the PSO, so that the diagnostic performance is improved.

The invention discloses a method for making a coarse cereal food, and the method comprises the following specific steps: 1) adding a quality modifying agent containing transglutaminase into coarse cereal powder in a certain proportion and mixing; 2) kneading dough; 3) performing heat preservation for certain time; and 4) preparing a finished product. By adopting the modifying agent, the processing performance of the coarse cereal powder dough can be improved, the network structure of the coarse cereal powder can be improved, the volume of the coarse cereal food can be improved, the side wall of coarse cereal bread can be reinforced, the slicing performance can be improved, the fermentation time can be shortened, the water absorption can be increased, dissolving-out substances can be reduced during stewing, and frost cracking can be prevented during freezing.

The invention discloses a method for identifying wire fracture damage of a steel wire rope based on wavelet transformation and a neural network, and belongs to the field of civil engineering structuredetection. The method comprises the following steps: firstly, detecting a steel wire rope with an unknown number of fractured wires by using a steel wire rope detection instrument, extracting a detected leakage flux signal and recording the parameters such as the diameter of the steel wire rope and the operation speed of the steel wire rope detection instrument; then, carrying out continuous wavelet transformation on the extracted and detected leakage flux signal by using multi-resolution characteristics and singularity detection capacity of wavelet transformation, and recording a wavelet coefficient peak-to-peak value after transformation; then, guiding the characteristic values such as recorded peak-to-peak value and the diameter of the steel wire rope into the neural network which hasbeen trained; and finally, obtaining parameters output by the neural network so as to complete quantitative damage identification for the number of fractured wires of the steel wire rope. According tothe method, quantitative identification for the number of fractured wires of the steel wire rope is realized, and the method has relatively high practicability.

The invention discloses a basalt fiber thermal insulation decorative integrated plate and a preparation method thereof, belonging to the technical field of building materials. According to the invention, carbon fiber and basalt fiber are taken as framework structural materials, modified SiO2 aerogel is taken as a heat insulation functional material, and inorganic fillers, adhesives and additives are combined to form the basalt fiber heat insulation decorative integrated plate. The basalt fiber thermal insulation decorative integrated plate can be used below 650 DEG C, has a thermal conductivity of about 0.02 W / (mK), is three to five times that of the traditional inorganic thermal insulation material, and can greatly improve the thermal insulation effect and high-temperature resistance.

This invention discloses a method for preparating a kind of inflatable vermiculite / poly-potassium acrylate-acrylamide super water-retention property water -absorbent composite material, which is made by graft copolymerizing acrylic acid, acrylamide and inflatable vermiculite powder. In this invention, inflatable vermiculite powder takes a art of 50%, whose distilled water absorption could amount to 850 times. The absorbency of inflatable vermiculite / poly-potassium acrylate-acrylamide super water -absorbent composite material is a little lower than that of pure organic resin polymer(it can fulfill normal using requirements), while its manufacturing cost is apparently less than that of pure organic resin polymer. Also it has very good water-absorbency and water-retention property, so it has very extensive use in agriculture and forestry, plant planting, soil improvement, food hygiene, buiding, petroleum chemical industry, and so on.

The invention discloses a high-impact resistance surface functional group treated carbon nanotube modified reactive powder concrete. The high-impact resistance surface functional group treated carbonnanotube modified reactive powder concrete is characterized in that the raw materials of the high-impact resistance surface functional group treated carbon nanotube modified reactive powder concrete comprise cement, functionalized multi-walled carbon nanotubes, silica fume, fly ash, sands, a water reducer and water; and the functionalized multi-walled carbon nanotubes are carboxylated multi-walledcarbon nanotubes or hydroxylated multi-walled carbon nanotubes. The functionalized multi-walled carbon nanotubes can significantly enhance the dynamic compressive strength, dynamic peak strain and impact toughness of the reactive powder concrete. When the strain rate is in a range of 200-800 / s and the dosage of the multi-wall carbon nanotubes is 0.25-0.5%, the dynamic compressive strength of concrete can be increased by 71.3%, the dynamic peak strain can be increased by 101. 1%, and the impact toughness can be increased by 100.8%.

The invention provides a high humidity resisting modified starch-base biodegradable film and a preparation method thereof. The invention aims to solve the problem of poor mechanical property of the present starch-base biodegradable film under a high humidity environment. A nanometer biodegradable film is composed of starch, PVA, water-based bentonite and glycerin. The preparation method comprises the following steps: pasting starch, and then performing moderate enzymolysis modification on the starch; mixing with a PVA aqueous solution, a water-based bentonite dispersion liquid and glycerin; and cooling and then pouring into a mold, thereby obtaining the high humidity resisting modified starch-base biodegradable film. The biodegradable film provided by the invention is simple in preparation process; under the condition of high humidity, the biodegradable film has an excellent stretching property; and the cost is guaranteed and meanwhile the application scope can be expanded.

The invention relates to a coating defect detection method based on deep learning, and the method specifically comprises the steps: 1, selecting different types of defect image blocks and defect-freeimage blocks as a training sample set; 2, constructing a convolutional neural network initial model; 3, carrying out data set training and model optimization; and 4, detecting and identifying the defects of the glued surface on line by using an optimized deep learning algorithm to realize automatic classification of the defects. According to the method, the defect image is directly input into thedeep convolutional network to construct a multi-layer neural network, layer-by-layer feature extraction is performed on the image, advanced features reflecting implicit in image data can be accuratelylearned under training of a large number of training data sets, and the network structure is optimized and trained to obtain an optimal parameter value; the problem of multi-defect type detection andrecognition in the coating melt-blowing process is solved.

A face image age recognition method based on an improved ensemble learning strategy comprises the following steps that (1) a plurality of weak classifiers need to be obtained in an ensemble learning model, each weak classifier can independently achieve prediction estimation on an input object, and a prediction model containing the weak classifiers is constructed; 2) a plurality of weak classifiersobtained based on the DCNN and a strong classifier obtained through integration all adopt softmax classifiers; 3) aan improved integrated learning strategy is adopted, firstly, following a combination method of a voting principle, and the credibility of each weak classifier'opinion ' is controlled by using a set threshold value T; Then, when the trust degree of the weak classifiers is generally low, a voting combination method is abandoned, and the confidence coefficient ai of each weak classifier is calculated to serve as the respective weight value; And finally, a probability distribution array of the strong classifier is obtained by using a weighted combination method, and a classification label corresponding to the maximum component of the strong classifier is taken as a final prediction result. . According to the invention, the accuracy is obviously improved;.

The invention discloses a computer cable electromagnetic leakage characteristic analysis method. The method comprises the following steps of identifying a leakage source, judging whether information leakage exists or not and analyzing leakage information content; under the condition that the electromagnetic signal characteristics are unknown, the electromagnetic signal can be analyzed, the leakage source of the electromagnetic signal and the leakage information hidden in the leakage source are gradually identified, the layered detection strategy of the electromagnetic information leakage is realized under the unified algorithm and technical framework, and the detection accuracy is improved. Compared with a traditional method for analyzing electromagnetic leakage information through operation experience and priori knowledge, the electromagnetic leakage information analysis efficiency is improved, under the condition that the signal-to-noise ratio is low, features in signals can be accurately extracted to judge a leakage source, the speed of solving the optimal value through gradient descent in the model is higher by adding a batch standardization layer, and the method has the advantages of being high in accuracy and high in reliability. And meanwhile, the overfitting problem is more effectively avoided, and the integrated learning method is adopted, so that the electromagnetic leakage information detection efficiency and accuracy can be relatively high under the condition of small samples.

The invention discloses an Internet of Things system and an equipment authentication method and device thereof, electronic equipment and a computer readable storage medium, the method is applied to aserver connected with a core switch, and the method comprises the following steps: acquiring equipment information of each piece of terminal equipment accessed to an access switch; after an equipmentinformation auditing result of an administrator is received, storing the equipment information of the terminal equipment passing the auditing into a user database; after an authentication request, sent by the access switch, of the terminal equipment is received, searching a user database for equipment information of the terminal equipment; and performing authentication response according to the search result. According to the invention, the network structure of the Internet of Things system is reasonably optimized, the authentication security level is effectively improved, the number requirement of servers is reduced, the equipment cost and configuration workload are greatly reduced, the equipment utilization rate is improved, the auditing operation flow of access authentication is simplified, and the working efficiency is improved.

The invention discloses a hydraulic oil filter paper for realizing high-efficiency oil-water separation, which is prepared from the following raw materials in parts by weight: 3-5 mm alkali-free ultrafine glass fiber 50-55, bleached coniferous wood pulp 55-60, distilled water in an appropriate amount, Anionic polyacrylamide 0.8-1, an appropriate amount of concentrated sulfuric acid with a concentration of 98wt%, nano-titanium dioxide 10-12, a solid content of 30% iron ferric oxide colloid solution 20-25, an appropriate amount of absolute ethanol, and a solid content of 40wt% fluorine Silicon emulsion 60-65, dispersant NNO0.8-1, zeolite molecular sieve 3-4, sodium polystyrene sulfonate 0.2-0.3, peat soil 2-3, sodium bicarbonate 0.4-0.5, microcrystalline paraffin 0.8-1. The filter paper of the invention has good adsorption efficiency, realizes oil-water separation, improves the clarity of hydraulic oil, and prolongs the service life.