242results about How to "Enough precision" patented technology

The invention discloses a three-dimensional gaze estimation method based on irises and pupils. The method comprises the steps that the characteristic that the centers of the pupils and the centers of the irises do not share the same point in the three-dimensional space is utilized, the centers of the pupils and the centers of the irises form the direction of the optical axis, the visual axis of eyes and the optical axis intersect with the center of curvature of the corneas, the direction of the visual axis of the eyes is obtained through calculation of the direction of the optical axis by calculating the deflection angle between the optical axis and the visual axis of the eyes, and in other words, the three-dimensional gazing direction of the human eyes is obtained. The novel high-precision model method is provided for three-dimensional gaze estimation.

The invention discloses an intrusion detection method based on traffic visualization and a machine learning algorithm. The intrusion detection method comprises the following steps: S1, capturing traffic by using high-speed capture equipment; s2, sending the flow which cannot be identified by the intruder database and the packet header information required by the intruder database to a data processing layer for data processing; s3, converting the received flow subjected to data processing into a grey-scale map; s4, based on semi-supervised learning, clustering the grey-scale map by using a K-means algorithm and classifying the gray-scale images of each cluster after clustering by using a CNN, and judging whether unknown invasion occurs or not based on an entropy theory and a classificationresult; s5, according to a classification result, based on an antibody theory in an AIS algorithm, purifying the specific attack by adopting a decision tree algorithm to obtain a detection result; themethod solves the problems that in the prior art, each attack cannot be accurately detected, the network attack cannot be detected in real time, the intrusion system establishment speed is low, feature extraction is complex, and the resource occupancy rate cost is high.

The invention discloses a dynamic simulation test system for ensuring running safety of a train on a railway bridge under an earthquake. The dynamic simulation test system comprises a bridge model and a train model, wherein a supporting frame component for supporting the bridge model and adjusting an inclination angle and a horizontal position of the bridge model is arranged at the bottom of the bridge model; track boards are paved on the bridge surface of the bridge model; train tracks for guiding a running route of the train model and central tracks for preventing the train model from falling off are paved on the track boards; the train tracks are parallel to the central tracks; the bridge model comprises a first supporting section for adjusting a preset speed of the train model through the inclination angle, a test section for performing a train-bridge coupling vibration test and a second supporting section for stopping the running of the train model according to the inclination angle; the first supporting section and the second supporting section are respectively connected to two ends of the test section. The weak action of the bridge, which is caused by the upper train, and strong action of bridge vibration, which is caused by a vibration table, are mutually coupled to realize a train and earthquake mixing simulation system.

The invention provides a modeling method for a generator excited system which comprises the following steps: (1) obtaining the idle load characteristic of the generator, the idle load and the load characteristics of an exciting dynamo, the time constants of the generator and the exciting dynamo, the phase position and frequency characteristic of the generator excited system and the damping effects of low frequency oscillation by experiments; (2) building a field system model; (3) analyzing the obtained basic data to obtain normalized model parameters, building an original simulation model in a Matlab emulator to carry out step response simulation and carrying out corresponding corrections of the parameters in the model according to simulation calculation and field experiment errors until the errors meet the requirements; (4) transforming the parameters into practically useful model parameters in a power system analyzing program and carrying out further correction of the model parameters in the program according to the specific characteristics of the original model. The model parameters of the field system obtained by the invention have the advantage of enough precision, which can really reflect the performance indexes of the field system.

The invention provides an electric current detection device, an electric current detection signal comparing unit, an electric current detection method and an electric current detection signal comparing method. The electric current detection device comprises an electric current detecting circuit, a control circuit, and a compensation circuit, wherein the electric current detecting circuit is used for detecting an electric current signal of a switching circuit so as to generate an electric current detecting signal, the control circuit is used for outputting an control signal, the compensation circuit is used for carrying out compensation on the electric current detecting signal according to the control signal; and the electric current detecting signal after compensation is compared with a constant current reference signal so as to output a constant current control signal. In addition, in the invention, on the other hand, the compensation circuit is set as compensation for the constant current reference signal, and the electric current detecting signal is compared with the constant current reference signal after compensation so as to output the constant current control signal. The electric current detection device, the electric current detection signal comparing unit, the electric current detection method and the electric current detection signal comparing method provided by the invention can apply to compensate an electric current detecting error caused by the switching circuit and the electric current detecting circuit, thereby obtaining the electric current detecting signal really reflecting the current output and having enough precision.

The invention relates to a water cooling wall on-line failure early warning method, which comprises the following steps: water power calculation, metal wall temperature calculation, stress analysis, fatigue life consumption analysis, creeping life consumption analysis and entire life consumption analysis. The method calculates the parameters of the flow rate of all pipes of a pipe set, inlet and outlet pressure, wall temperature values in a furnace of each section of the pipes along the length, stress change amplitude values, and the like through a flow and heat conduction rule in real time on line. A small amount of measuring points which are distributed in the pipe sections of each radiation area provided with flow and heat conduction representation and the existing furnace outside inlet and outlet measuring points are combined as calculating and checking points, and the calculating and checking points are input into a computer to calculate and process so as to judge whether unsafe factors of unstable water power, overhigh temperature, fatigue creeping destroy, and the like exist or not. A calculating result displays and records relevant data in real time, and is used as a reference for judging whether the wall cooling wall is safely operated or not so as to achieve the optimal operation of equipment, arrange the examination and maintenance management according to the state, and fully implement equipment state overhaul.

The invention discloses a magnetic core loss separation calculating method. The method includes the following steps: under the condition of the given magnetic flux density variable quantity, , forming a magnet loss construction separation equation set to separate magnetic hysteresis loss from vortex loss by using loss data of any two working points with different frequency under sinusoidal excitation, simulating corresponding relation of resistance rate of magnetic core materials and frequency of the magnetic core materials through a linear interpolation method, and gathering the additional vortex loss of the rest loss to a classical conduction vortex loss. The magnetic core loss separation calculating method has the advantages of being clear in physical concepts, simple in separation calculation process, capable of effectively separating loss components of the magnetic core without dependence on a special instrument, and being widely applied to simulation software to simulate size of loss of the magnetic core materials in power electronic application under different frequency and magnetic induction intensity.

The embodiment of the invention discloses a deep learning model training method and device, training equipment and a storage medium. The deep learning model training method comprises the following steps: training a deep learning model by using a training image to obtain training characteristics output by the deep learning model; carrying out conversion processing on the training features by usingan auxiliary training module to obtain conversion features; Determining a loss value based on the conversion feature; And determining whether to continue to train the deep learning model based on theloss value.

The invention discloses a high precise stereo measuring method for seamless stereo mould based on the digital orthogonal image and stereo assistant image. The method calculates inversely the measured isonym image point on the seamless stereo mould into the original stereo image pair of the image point. Then, it calculates the three-dimension coordinate precisely through spatial front crossing; the accuracy doesn't be affected by digital ground mould difference in the producing process of the stereo mould, in condition of that the resolution of the seamless stereo mould digital orthogonal image and the stereo assistant image and the aviation photo's resolution is consistent, the accuracy can reach the one with original stereo image measuring.

The invention provides a micro optical DC/AC electric-field sensor for eliminating the drift error of electric charges, capable of being used for measuring the voltage of a high-voltage DC transmission line by an optical method. Two transparent electrodes formed on an electro-optical crystal light-pass end surface used in an optical sensing module are electrically and compactly connected with two output terminals of an optical relay respectively. A secondary light source transmits the light beam that is open and is cut off periodically to a light beam collecting window of an the relay by auxiliary optical fiber under the coordinated control of a signal processing and control module, and the light beam that is open and is cut off periodically is used for controlling the breakover and the break of a circuit at the output end of the optical relay; the modulation to the DC electric field to be tested which is externally added on an electro-optical crystal can be realized, so that the DC electric field to be tested is periodically applied and eliminated on the electro-optical crystal. The error caused by space charge drift in the electro-optical crystal used when the DC electric field is measured can be controlled within bounds by the micro optical DC/AC electric-field sensor, thus realizing the long-term, stable, reliable and precise measurement on the DC electric field with the optical method.

A soft method for measuring oxygen content in smoke discharged from large capacity boiler includes checking accuracy of original signal, confirming heat ratio of coal to air, calculating heat amount generated by fuel burning, calculating dynamic heat quantity and calculating oxygen content in smoke discharged from said boiler.

Provided is a high-accuracy fast neutron reactor assembly few-group cross section obtaining method. The fine-group cross section of a fast neutron reactor assembly is calculated in the mode that point cross section data and multi-group data are combined, online calculation of an elastic scattering matrix is performed, a neutron transport equation is solved by using the fine-group cross section to obtain fine-group neutron-flux density, energy groups are merged on the basis, and accordingly few-group parameters of the fast reactor assembly are obtained. Due to the fact that the method directly uses the point cross section data, the elastic scattering resonance effect of a medium mass nuclide and the resonance interference effects of all nuclides are accurately considered. The neutron-flux density of a real system is adopted in the energy group merging process, and merging results are more accurate. The overall errors of fast reactor assembly few-group parameters calculated by adopting the method are within 1% compared with a reference value, and the method has higher accuracy.

The invention relates to the establishment of ship motion mathematic models for navigation simulators, in particular to the establishment of a KaMeWa type water-jet hydrofoil motion mathematic model for navigation simulator. Firstly, a calculation method for a hydrofoil lift force, a hydrofoil attack angle and a hydrofoil inertia force considering free surface effect of hydrofoil is established according to the hydrodynamic characteristic of hydrofoil, then a calculation method for the wave surface at hydrofoil is established considering the interference of outside wind and waves, and a ship body parameter of buoyant center position, buoyancy, water plane area and waterline length are calculated in real time in the condition of variable draft. The model also comprises a method for calculating the hydrodynamic force of a KaMeWa type water-jet unit of an inverted navigation bailing bucket. By adopting the mathematic model, the requirements of the navigation simulator operating the KaMeWa type water-jet hydrofoil in the conditions of various operation situations and interference of outside wind and wave in training and scientific research can be met.

The invention discloses a blade of a high-power horizontal shaft ocean current energy electric generator. The cross section of a metal girder of the blade is shaped like the pictures in the specification; a metal blade tip is welded to the top end of the metal girder; a first net cutter is fixed to the metal blade tip; a metal blade root is welded to the bottom end of the metal girder and is a flange or a screwed hole which is connected with a hub of the ocean current energy electric generator; a first fiber fabric which is smeared or soaked by a first mixed resin wraps and is solidified on the outer side of the metal girder; an outer covering is stuck to the outer side surface of the metal girder; an inner covering is stuck to the inner side surface of the metal girder; gaps between the front side surface and the back side surface of the metal girder and the outer covering as well as the inner covering are filled with foam filling cores; a second net cutter is inserted to the foam filling core at the lower end of the metal girder; a front edge metal wrapping sheet is stuck to one side of the rotating direction of the blade; an anti-pollution coating is sprayed onto the surface of the blade. The blade has the advantages of tide load impact resistance, high strength, high rigidity, gravel collision friction resistance, abrasion resistance, blade tip cavitation corrosion resistance, fishing net winding resistance and marine organism adhesion.

Tool for cutting materials with a rotatable body having a rotation axis and cutting edges for cutting the material during movement of the body in a first direction parallel to the rotation axis. In order to stabilise the tool during cutting the cutting edges comprise inner cutting edges laying on a first surface of revolution which is in the first direction higher at a larger diameter and lower at a smaller diameter. The invention includes methods for using the tool.

The invention provides a modeling method of using a membrane humidifier to realize fuel-cell intake-air humidification. A transient simulation model of the one-dimensional flat membrane humidifier isestablished. Governing equations of the model include four conservation equations of a water content in membrane, water vapor in a flow channel, liquid water in a wet side flow channel and energy. High-heat high-humidity exhaust gas of a fuel cell stack is used to humidify desiccated low-temperature reaction gas at a humidification inlet, and a water-heat coupling transfer process inside the humidifier, phase change processes among the water vapor, the liquid water and membrane water and a combined humidification function of water and gas are fully considered. The model is based on a display format updating algorithm, the governing equations are solved at a center of each layer, calculation efficiency is high, and sufficient model accuracy can be guaranteed. The membrane humidifier model can realize coupling with the electricity stack, and a perfect humidification system is constructed through setting of boundary conditions to be used as a numerical simulation tool of heat and mass transfer analysis, structure optimization and type selection of a system, and has very high guiding significance for actual product development.

The invention discloses a photo current monitoring device used in high-speed fiber-optical communication, which is characterized by comprising a mirror image circuit and a clamp module, wherein the mirror image circuit consists of two parallel and consistent branch circuits; each branch circuit sequentially consists of a source electrode and a drain electrode of an adjusting pipe and two series resistors; the mirror image circuit further comprises an adjusting module and a clamp module; the adjusting module is used for forming a voltage-stabilizing circuit with the adjusting pipe; and the clamp module is used for clamping voltages at two ends of the mirror image circuit. The circuit integrates voltage-stabilizing bias and mirror image monitoring, thus providing a voltage-stabilizing circuit and a current source for a photodiode and ensuring the working state of the photodiode on one hand, and outputting monitoring current not influencing the photodiode so as to ensure the consistency and stability of the monitoring precision and the mirror image circuit, being capable of realizing muA level current monitoring, simultaneously ensuring enough precision of the photo current detectionin a wide range, and having small influence of a monitoring output end on transimpedance amplifier noise and bandwidth.