10results about How to "Small gradient" patented technology

The invention belongs to the technical field of reliability-based design optimization, and particularly relates to a reliability design optimization method combining transfer learning and a physical information neural network. Comprising the following steps: constructing a physical information neural network model by adopting training data of a source domain, and evaluating the reliability of the source domain. A physical information neural network model constructed by adopting training data of a source domain is updated by utilizing a layer freezing and fine tuning technology in transfer learning, so that the reliability of a target domain with different statistical attributes is evaluated, the calculation overhead is saved, and the calculation efficiency is improved. A first-order scoring function is utilized to carry out random sensitivity analysis on the failure probability so as to ensure normal operation of design optimization. And finally, providing estimation reliability, an objective function value and corresponding sensitivity information for an optimizer to obtain a corresponding updated design point. And repeating the iterative design process until the optimal design is realized. The training speed of the physical information neural network model is increased, and the precision of the physical information neural network model is improved.

The invention provides a stator assembly, an axial magnetic flux motor, an electric driving device, a system and electric equipment.The stator assembly comprises two iron cores, the air gap sides of the two iron cores are oppositely arranged or oppositely arranged, and the air gap sides of the iron cores are provided with a plurality of winding grooves arranged around the central axis of the stator assembly; the projection of the iron core in the axial direction of the stator assembly is a first projection, and in the first projection, the length direction of the winding grooves and the radial direction of the iron core are relatively inclined. According to the stator assembly provided by the embodiment of the invention, the inclination amplitudes of the winding grooves of the two iron cores are the same and the inclination directions are opposite, so that the magnetic flux can be alternately transmitted between the two iron cores, the armature reaction of the axial magnetic flux motor is effectively weakened, the eddy current loss of the axial magnetic flux motor is effectively inhibited, and the service life of the axial magnetic flux motor is prolonged. Compared with a mode that two windings adopt different winding structures, the stator assembly has the advantages that the two windings of the stator assembly can adopt the same winding structure, so that the manufacturing and assembling processes of the axial magnetic flux motor are effectively simplified.

The utility model discloses a lens, an automobile lamp optical system and an automobile lamp assembly, and relates to the technical field of automobile lighting. The lens comprises a lens body, the surface of the lens body is sequentially divided into a first area, a second area and a third area in the longitudinal direction, the second area is located in the middle of the lens body, and the first area, the second area and the third area are provided with a first grid pattern, a second grid pattern and a third grid pattern respectively. The arch height of the second grid pattern is larger than that of the first grid pattern and that of the third grid pattern. The lens, the automobile lamp optical system and the automobile lamp assembly can solve the problem that the existing high and low beam automobile lamp optical system cannot meet the requirements of reducing the low beam light-dark cut-off line gradient and not influencing the high beam light type maximum illumination value and the light type energy distribution and shape at the same time.

The invention provides a model training method and device, a depth estimation method and device, electronic equipment and a computer readable storage medium. The method comprises the following steps: acquiring a plurality of image samples and a depth image of each image sample; for each depth image, constructing Gaussian distribution corresponding to each pixel point according to the depth value of each pixel point in the depth image; performing depth estimation on the image sample by using a depth estimation model to obtain depth probability distribution of each pixel point; determining a loss value based on the depth probability distribution and the Gaussian distribution; and obtaining a trained depth estimation model based on the loss value. Through the method and the device, the unbiased transformation process of the loss function can be constructed through Gaussian distribution, the loss function of the model is improved, the convergence speed of the model can be improved, and thus the model training efficiency is improved.

The invention discloses a semiconductor structure and a forming method thereof. The semiconductor structure comprises a substrate; the gate structure is formed on the surface of the substrate; the pair of side walls are formed on the side walls on the two sides of the gate structure respectively; the source electrode region and the drain electrode region are respectively formed on the substrate at two sides of the gate electrode structure; the metal silicide layer is formed on the surfaces of the gate structure, the source region and the drain region; wherein for any one of the source electrode region and the drain electrode region, a non-zero gap is formed between the region and the side wall, and / or the distance from a metal silicide layer formed on the surface of the region to the gate structure is greater than the width of the side wall. According to the scheme, the voltage resistance and the electric leakage performance of the device can be reliably improved on the premise that the threshold voltage is hardly influenced.

This utility model discloses a hydrogen-rich water device with adjustable hydrogen-water mixing, including a pure water tank, a pure water circulation pump, an electrolyzer, a gas-water separator, and a buffer tank. The electrolyzer and gas-water separator are equipped with an anode, a cathode, and a proton exchange membrane. A motor mounted on the buffer tank drives a high-pressure hydrogen dissolving rod to move left and right. Components such as the insertion plate and assembly plate have specific connections and are equipped with sealing sleeves to prevent interference. This device has dual channels for electrolyzing water and drinking water, allowing the human body to drink different water qualities through a series of processes, improving its practicality. Furthermore, the reciprocating motion of the high-pressure hydrogen dissolving rod driven by the motor increases the contact area between hydrogen and water, forming a flow path and vortex, improving the hydrogen dissolution speed and efficiency, resulting in a more uniform hydrogen distribution, reducing concentration gradients, and improving the quality of the hydrogen-rich water and the overall quality of the device.

The utility model relates to the technical field of SMP connector sintering, in particular to a sum-difference network SMP sintering tool. According to the technical scheme, the sum-difference network SMP sintering tool comprises a substrate, an SMP connector, a positioning pin, a clamping groove and a protruding block, the clamping groove is formed in the SMP connector, the positioning pin is placed in the clamping groove, and the SMP connector is placed on the substrate in a limited mode through the positioning pin. According to the utility model, the perpendicularity and the height of six SMP connectors can be solved when the SMP connectors are sintered at the same time.

The invention relates to the technical field of concrete, and particularly discloses underground space anti-crack and anti-corrosion mass concrete and a preparation method thereof. The invention discloses anti-crack and anti-corrosion mass concrete for underground space. The anti-crack and anti-corrosion mass concrete is prepared from 320 to 335 parts of cement, 76 to 88 parts of fly ash, 1025 to 1050 parts of coarse aggregate, 670 to 685 parts of fine aggregate, 11 to 15 parts of carbon fiber, 14 to 17 parts of graphite particles, 4 to 6 parts of anti-permeability agent, 2 to 3 parts of water reducing agent and 155 to 170 parts of water, the anti-permeability agent is prepared by mixing polyethylene glycol distearate, diethanolisopropanolamine and D-mannitol according to the mass ratio of (2 to 4): (1 to 2): (1 to 2). The mass concrete prepared by the invention has good crack resistance and erosion resistance, can effectively control the generation of cracks under the condition of not adding an expanding agent, and can be widely applied to underground space engineering such as underground stations, tunnels, basements and the like.

A system and method for welding carbon fiber reinforced thermoplastic workpieces (10, 12) comprising upper and lower tools (1, 2), each tool having a holder (4.1, 2.1) for holding the respective workpiece (10, 12) in a cantilevered manner during heating (6), and a pusher (4.2, 2.2) movable relative to the holder (4.1, 2.1) for pushing the workpieces (10, 12) once the workpieces are brought together by the tools (1, 2). The heated regions (10.1, 12.1) of the workpiece are pressed against each other. This arrangement allows the heated ends (10.1, 12.1) to be surrounded by air, providing a smaller, more uniform thermal gradient over the thickness of the respective workpiece (10, 12). This also helps to reduce the heat dissipation from the workpiece (10, 12) to the tool (1, 2), in particular in materials with high thermal conductivity. Ideally, heating is provided by a movable infrared heater (6) cooperating with the shielding screen (8).

The present disclosure relates to a separator for an electrochemical device and a method for manufacturing the same. Particularly, the drying rate of a composition for forming a second porous coating layer is controlled to be higher than the drying rate of a composition for forming a first porous coating layer. In this manner, it is possible to improve the interfacial adhesion between an electrode and the separator, while ensuring adhesion strength between a porous polymer substrate and a porous coating layer.