46results about How to "Avoid tuning" patented technology

The invention discloses a multi-target tracking method based on a recurrent neural network, comprising the following steps: building a monitoring video data set marked with each frame of pedestrian location; manually expanding the monitoring video data set marked with each frame of pedestrian location to get training set samples; grouping the training set samples to get multiple training groups; building a multi-target tracking network; inputting the training groups in sequences to the multi-target tracking network for training; and inputting video data to be tested to the trained multi-target tracking network and performing forward propagation to get the moving trajectories of multiple targets. According to the invention, a proposed network model is trained in an end-to-end manner using raw data and a large amount of data obtained through artificial expansion, complex tasks such as data association and track estimation are completed under a unified neural network architecture, and the moving trajectories of targets can be tracked effectively under different directions, light conditions, deformations and other complex environments.

The invention discloses a fuse trimming and adjusting circuit comprising a switch control module, a trimming and adjusting value loading module, a fuse fusing control module and a trimming and adjusting module. The trimming and adjusting module comprises a PMOS transistor, a first resistor, a fuse, an NMOS transistor, a second resistor, and a D trigger. The source electrode of the PMOS transistor is connected with a voltage-stabilization power source; the grid electrode of the PMOS transistor is connected with the switch control module; the drain electrode of the PMOS transistor is connected with one end of the first resistor; the other end of the first resistor is connected with one end of the fuse; the other end of the fuse is connected with the source electrode of the NMOS transistor; the grid electrode of the NMOS transistor is connected with the fuse fusing control module; the drain electrode of the NMOS transistor is grounded; one end of the second resistor is connected with the source electrode of the NMOS transistor; the other end of the second resistor is grounded; the CP port of the D trigger is connected with the trimming and adjusting value loading module; and the D port of the D trigger is connected with the source electrode of the NMOS transistor. With the fuse trimming and adjusting circuit, a wafer can be trimmed and adjusted after being packaged, thereby reducing cost and improving trimming and adjusting precision.

The invention relates to the field of unmanned control methods, in particular to a course control system and method for an unmanned surface vehicle. The control system comprises an upper computer, a transmission station and a lower computer. The control method includes the set steps that an expected course is input in a human-computer interaction interface of an upper computer operating deck and transmitted to the lower computer through the transmission station; information is acquired and transmitted, wherein unmanned surface vehicle comprehensive information acquired by the lower computer is transmitted into the upper computer through the transmission station; and control is conducted, wherein the upper computer conducts calculation and processing according to the received unmanned surface vehicle comprehensive information, a rudder angle delta is output and transmitted to the lower computer, and the rudder direction is automatically changed through a course control device, so that the course of the unmanned surface vehicle is adjusted. The system and method have the beneficial effects that maneuverability is high, robustness is strong, a stepping motor and a bevel gear wheel are simultaneously adopted for driving the steering course control device, the control accuracy is higher, steering is more accurate, and the application prospect is good.

The invention relates to a logic operation-based distributed feeder automatization fault judgment method. The method is characterized in that in combination with the features of a power distribution network, an open-loop operating mode and a closed-loop operating mode of the network adopt a unified fault judgment rule, so as to avoid parameter adjustment due to the changing of operating modes. Through a fixed wire connection manner, the direction of the flow passing through a switch is set, and the flow and fault current passing through the switch are converted into unified analog quantity, and the control state of the switch can be obtained according to the logic operation of adjacent switch groups. A voltage-loss reclosing function on one side of the switch is added, so as to effectively avoid the influence on a grid, which is caused by transient fault. The method is concise, convenient and strong in adaptability, and can meet the actual requirement of feeder automatization.

A variable speed wind turbine having a doubly fed induction generator (DFIG), includes an exciter machine mechanically coupled to the DFIG and a power converter placed between a rotor of the DFIG and the exciter machine. Thus, the power converter is not directly connected to the grid avoiding the introduction of undesired harmonic distortion and achieving a better power quality fed into the utility grid. Moreover, the variable speed wind turbine includes a power control and a pitch regulation.

The invention discloses a method for suppressing subsynchronous oscillation of a doubly-fed wind power plant based on a stator side analog resistor. The method comprises the following steps: firstly,establishing a system model of the doubly-fed wind power plant with series compensation capacitor grid connection; secondly, deducing an equivalent impedance model of the doubly-fed fan grid-connectedsystem in combination with a control strategy of a rotor side converter, and analyzing damping characteristics of the system; determining the value of a stator side analog resistor according to the analysis result of the damping characteristic of the doubly-fed fan grid-connected system so as to compensate the negative damping effect of the system; and finally, applying a control strategy based on the stator side analog resistor to control of a doubly-fed fan stator side converter, and completing suppression of subsynchronous oscillation of the system. The subsynchronous oscillation phenomenon in the wind power plant grid-connected system can be effectively inhibited, the stability of the system is improved, any additional device is not needed, the cost is saved, and the practical engineering application is facilitated.

The invention discloses a substation automation information parameter setting method comprising the following steps of: formulating or maintaining a rule template library according to an information standard of a dispatching center; importing a 61850 static model configuration text and a rule template; generating a primary system model and a secondary system model of a whole substation according to the static model and the rule template; generating an information parameter demand schedule according to the rule template; matching measuring points according to the incidence relation and the naming notation between primary equipment and secondary equipment to form a final information parameter setting schedule; and importing the formed information parameter setting schedule into a substation remote control gateway device and a master station of the dispatching center. The method can be used for overcoming the defects of low efficiency and no standard of the traditional automatic information parameter setting method and mapping an information standard into an actual measuring point of an automatic device in the substation to form a substation remote control information list used by substation and master station ends.

The invention provides an optical system with a visibility adjustment function. The system comprises an image source for generating image light, a first lens group, a liquid lens with a zooming function, a second lens group, and a free-form-surface prism. The image light sequentially passes through the above optical members and reaches an exit pupil plane to enter a human eye to carry out imaging,and the visibility adjustment function is realized. The visibility adjustment requirement of a wearer is satisfied, and the wearing convenience and comfort are improved. The liquid lens and the likeare used to realize visibility adjustment, the design and adjustment of a complex axial mechanical zoom structure are avoided, combined with a visual magnification element of a free-form-surface prismmode, in the design of an optical imaging system, an optical imaging system with small distortion, high image quality and a compact structure can be provided, and the assembly error of the optical system is reduced.

The invention discloses a super capacitor energy storage device for assisting frequency modulation, and the device comprises a power grid, a three-phase transformer, a filter inductor and a bridge armmodule; the bridge arm module comprises a first bridge arm module, a second bridge arm module and a third bridge arm module, and the filter inductor comprises a first filter inductor, a second filterinductor and a third filter inductor. The invention further discloses a control method of the super-capacitor energy storage device for assisting frequency modulation. According to the super-capacitor energy storage device, the online switching and replacement of the super-capacitor sub-modules are achieved, and energy storage and overall integrated design of the converter are facilitated; the switching loss is reduced, the voltage-sharing process of the super-capacitor sub-module is simplified, and parameter setting of a large number of voltage-sharing loops is avoided; and a large number ofswitching devices are saved on the basis of ensuring that output waveforms meet power grid harmonic requirements.

The present invention relates to a broadband vibration suppression device using acoustic black hole characteristics, the device comprising: a vibration damping body including a uniform portion havinga constant thickness and a tapered portion integrally connected to the uniform portion, the tapered portion extending from a connection point in such a manner as to gradually decrease in thickness from the uniform portion to a predetermined thickness; and a damping layer attached to the tapered portion of the vibration damping body. The vibration suppression device can effectively suppress vibration and reduce noise on a broadband, and is simple in design.

The invention relates to a saturated switching control method for a three-phase parallel active power filter. The method comprises the following steps of: 1) establishing a switching system model according to each operation mode of the three-phase parallel active power filter; 2) constructing a common Lyapunov function to obtain state-feedback-based switching rules, wherein arg represents a subscript meeting an expression; 3) introducing an additional matrix G in case of saturated nonlinear constraints of the three-phase parallel active power filter, linearizing the saturated nonlinear constraints by using a convex domain method, and designing saturated switching rules capable of progressively stabilizing a saturated system by utilizing a Lyapunov function stability theory; and 4) compensating the harmonic current of a power grid by using the three-phase parallel active power filter under the action of the saturated switching rules.

The invention provides a modularization multi-level converter model prediction control method. The modularization multi-level converter model prediction control method includes the following steps: calculating a dq axis reference current according to a power command; conducting error correction on the dq axis reference current; calculating a cost equation through a phase current predicted value, and selecting the optimal phase voltage; calculating the cost equation through current predicted values inside bridge arms, and selecting the optimal compensation level; determining the number of submodules needing to be put into the bridge arms; calculating the cost equation according to a capacitor voltage predicted value of the submodules, determining the submodules put into the bridge arms; giving a trigger command that an upper switch tube is closed and a lower switch tube is opened to the submodules selected to be put in; giving a trigger command that the upper switch tube is opened and the lower switch tube is closed to the rest of the submodules. According to the modularization multi-level converter model prediction control method, the calculation amount of model prediction control in MMC control is reduced, the phase current control steady accuracy is improved through reference current correction, the overall structure is simple, and the modularization multi-level converter model prediction control method is more suitable for the characteristics of flexibility and expandability of an MMC.

The invention discloses a stock intelligent prediction method combined with a news text. Firstly, a news text is pretreated, Chinese word segmentation and stop words are filtered, and news text without time label is deleted; the prediction time length Delta t of the stock is determined, the news text is filtered and selected according to the time label of the news text; feature representation is performed on the selected news text, and a feature representation vector xt0 of the corresponding time is formed by the features of the news text and the stock data feature vector of the correspondingtime; a self-encoder deep learning network is constructed, the feature representation vector xt0 is input to the deep learning network from an encoder for compression and feature extraction, and a low-dimensional eigenvector new_xt0 is obtained; an ELM neural network model is constructed, the change degree of stock price is expressed quantitatively, and the target output value of ELM neural network model is determined; the parameters of ELM neural network model are optimized and a final prediction model is obtained. The invention solves the technical problem of low accuracy of stock prediction through the combination of a news event and historical market data.

The invention relates to an adjustable-wavelength emitter (100) for a TWDM-PON. The adjustable-wavelength emitter comprises a multi-mode emitter (110) outputting an optical signal containing multiple longitudinal modes, an optical annular device (120) having a first port used for receiving the optical signal of the multi-mode emitter, a second port used for outputting the optical signal and a third port, and a first micro ring resonator (130) having a first input end used for receiving the optical signal from the second port, a first output end used for outputting a first component of a resonance portion of the optical signal and a second output end used for outputting a second component of the resonance portion of the optical signal, wherein the second output end is connected with the third port of the optical annular device (120).

The invention discloses a method for suppressing subsynchronous oscillation of a doubly-fed wind power plant based on analog inductance. The method comprises the following steps: firstly, establishinga system model of the doubly-fed wind power plant with series compensation capacitor grid connection; then applying an analog inductance control strategy to the inner loop control of a doubly-fed fanrotor side controller to control the output voltage of the rotor; secondly, performing self-adaptive online setting on the analog inductance value by utilizing a PSO algorithm, and determining an optimal value of the analog inductance under a specific working condition; and finally, applying an analog inductance control strategy and a rotor side converter control strategy to the control of the rotor side of the doubly-fed fan, and completing the suppression of the subsynchronous oscillation of the system. The method improves the electrical damping in the subsynchronous frequency band, improves the stability of the system, and can meet the subsynchronous oscillation suppression requirements under multiple working conditions.

A group welding method for five plates on the roof of a CRH5 motor train unit is characterized by comprising the following steps: step one, a forward-assembly reversible deformation supporting block is pre-placed at the position of a middle top plate of the roof, wherein the height of the forward-assembly reversible deformation supporting block is 5 to 7mm; step two, the welding sequence is that first performing backward-assembly group welding and then performing forward-assembly group welding; when performing the backward-assembly group welding, two welding lines in the middle are welded first, and then two welding lines at two sides are welded; when performing the forward-assembly group welding, the two welding lines at the two sides are welded first and then the two welding lines in the middle are welded. According to the group welding method provided by the invention, the whole assembly and the whole welding of the five plates on the roof are realized, so that all butt welding lines are welded in the same working procedure, and the size is qualified and adjusting and repairing is avoided. The separation of the butt welding lines and lap welding lines of roof compounding is realized, the group welding time of subsequent whole roof compounding is reduced to 360 minutes from 960 minutes, the whole roof adjusting and repairing time is reduced to 45 minutes from 180 minutes, and the operation of staff is reduced to 2-riveting and 1-welding from 4-riveting and 2-welding, the assembly difficulties in the subsequent roof compounding, the labor intensity and the welding deformation are reduced, and the labor efficiency and product quality are improved.

The invention discloses a PI control loop performance evaluation method with the controller output constraint. In a control loop, the controller output difference constraint is introduced to limit the variation of a controlled object. Meanwhile, the fluctuation frequency and the fluctuation range of the controlled object are limited by restraining the output difference of a controller, so that the controlled object does not fluctuate violently. Meanwhile, the output difference of the loop is as small as possible, so that the interference of the external disturbance to the loop is reduced. According to the technical scheme of the invention, the optimization range of PI parameters is gradually expanded, and the region is gradually enlarged. A minimum output difference is found in the defined region. In this way, found PI parameters can be guaranteed to have good application prospect in the minimum output difference performance evaluation. Meanwhile, the optimization result is ensured not to enable the output variation of the controller to be too large. As a result, two optimization targets can be taken into consideration at the same time, and a powerful guarantee is provided for the efficient, green and low-energy-consumption production of enterprises.

The invention discloses an automatic hand bone segmentation method based on a template. At first, a bone age sample database is established. Then samples are trained and templates is created, in thesame period of time, templates are createdfor all segments of the bone respectively. Finally, hand bone segmentation is performed based on template. The method extracts intact bone shape features including epiphysis, and more bone development information is contained. At the same time, the method avoids falling into endless tuning parameters; And through the training set update method, the workload of manual annotation is effectively reduced. The invention solves the problem of hand bone segmentation and provides a powerful tool for subsequent bone age assessment.

The invention relates to a robot multi-joint self-adaptation compensation method based on a perceptron model and a stabilizing device. The method comprises the steps that firstly, based on a robot inertial sensor and a gait generator, error data are obtained; secondly, the error data are input into the pre-built perceptron model, the perceptron model updates a network weighting parameter based onan associative learning strategy, and compensation value is calculated and output; and thirdly, based on the compensation value, a robot is subjected to motion compensation, and the compensation valueoutput by the perceptron model comprises hip joint compensation value, knee joint compensation value and ankle joint compensation joint. Compared with the prior art, the associative learning strategyis adopted, the network weighting parameter is updated through a supervisory Hebb learning rule, robustness is improved, self-adaptation control is achieved, compensation amount is dispersed to all joints, such as the ankle joint, the knee joint and the hip joint, of a leg, the joint loads are reduced, and the service life of a robot is prolonged.

The invention provides a bran-new image retrieval framework based on parameter-free quantum estimation. According to the framework, images of a database are regarded as photons generated by a light source, and the relevance of the images of the database is estimated through querying a polarization filter during retrieval. The basic idea of the framework is as follows: firstly, images with relatively low relevance in the database are filtered out by pre-filtering, so as to reduce the computing cost; then, the relevance probability between the images with relatively high relevance and an input query image is computed by using a semi-supervised learning method; finally, the relevance of the images with higher rankings is optimized through parameter-free quantum estimation, so as to obtain a satisfactory retrieval result.

The invention discloses a method for adjusting and optimizing performance of a linux system, which comprises the steps of selecting a reference tuning linux operating system compared with a to-be-tuned linux operating system, installing two systems on the same server, testing the two systems by using the same performance testing tool to test performance of the two systems by using the same performance testing tool; screening out the test items of which the test scores of the to-be-tuned linux operating system are lower than the test scores of the reference tuned linux operating system, then excluding irrelevant items from the low-score test items, and finally excluding and optimizing the remaining low-score test items after the irrelevant items so that all software packages of the to-be-tuned linux operating system are prevented from being sequentially subjected to performance testing and tuning. Therefore, performance defects can be quickly found for optimization, and the technical problems that in the prior art, a software package influencing the performance of the linux system cannot be quickly found, and the performance test and performance optimization efficiency of the linux system is improved are solved.

The invention relates to the technical field of welding deformation control and discloses a framework welding deformation control method and a framework welding tool. The method comprises the following steps: performing structural simulation analysis on a framework to divide the framework into a plurality of welding compositions according to modules and determining the welding sequence of welded joints among all the welding compositions; and based on the welding sequence, welding the welded joints among all the welding compositions. The framework welding tool comprises a clamp body, and a middle support, a torsion-proof device, a cross beam tight-pressing device and a side beam tight-pressing device which are arranged on the clamp body, so that mounting and fixation of the framework are facilitated. Repairing of the framework can be avoided, the manufacturing period is greatly shortened and efficiency and quality are enhanced.

The invention discloses a training method for an end-to-end speech recognition model. The end-to-end speech recognition model includes an encoder and a decoder. The method includes: obtaining an acoustic model and a cross-entropy language model through pre-training; Model initialization of the encoder of the end-to-end speech recognition model; initialization of the decoder of the end-to-end speech recognition model according to the cross-entropy language model; training of the end-to-end speech recognition model after initialization. The multi-stage pre-training method avoids the long and slow learning phase in the early stage of model training, thereby greatly reducing the model training time. At the same time, this strategy has no hyperparameters to tune. Compared with the existing technology, it avoids a lot of tedious hyperparameter tuning.

The invention relates to the technical field of welding, and particularly relates to a welding method of a side beam assembly and a welding control system. The welding method of the side beam assemblycomprises the steps that structural simulation analysis is conducted on the side beam assembly, so that the side beam assembly is divided into a plurality of welding components according to modules,and the welding assembling sequence of all the welding components is obtained; and welding and assembling of all welding components are completed according to the welding assembling sequence. According to the method, the problem of deformation of side beam welding is solved, adjustment and repair of rail vehicle side beam welding are avoided, the production cost, especially the time cost, of the side beams is saved, time waste in a side beam manufacturing process is reduced, and therefore the production efficiency is improved, and the production and manufacturing period is greatly shortened; and in addition, by means of the method, the mechanical and flame adjusting and repairing process of the side beam assembly can be completely avoided after welding, so that it can be better proved thatthe side beam assembly manufactured through the method has higher product quality.

The invention discloses a portable thermal radiation absorption spectrum methane detector. The detector comprises a handle. The top end of the handle is fixedly connected with a cavity. One side of the handle is rotationally connected with a trigger. The side face of the end, close to the handle, of the cavity is fixedly connected with a control panel. And the end, away from the handle, of the cavity is fixedly connected with a lens, the interior of the handle is fixedly connected with a battery, the side, close to the battery, of the interior of the handle is fixedly connected with a PCBA plate, and the top end of the PCBA plate and the top end of the battery both penetrate through the handle and are located in the cavity. The invention relates to the technical field of methane detectiondevices. According to the portable thermal radiation absorption spectrum methane detector, an uncooled infrared thermopile is used for replacing a photoelectric avalanche diode or a photomultiplier ofa laser spectroscopy method to serve as an optical probe. The cost is remarkably reduced, the selectivity of the absorption wavelength is realized through mechanical replacement of the band-pass filter, the tuning of the light source is avoided, and finally, an additional light source is not needed.

The invention discloses a nonlinear control method for a layered access extra-high voltage direct current transmission system, comprising the following steps: according to equipment parameters and control system parameters of an extra-high voltage direct current transmission system of a receiving-end layered access alternating current power grid, establishing a state equation and an output equation of the extra-high voltage direct current transmission system of the receiving-end layered access alternating current power grid; selecting coordinate transformation meeting a diffeomorphism condition to carry out accurate linearization on the state equation; solving a Riccati equation corresponding to the linear system according to an optimization principle in the linear system; according to optimal pre-control of solving, forming an optimal control law of the layered access ultra-high voltage direct current transmission system, and optimizing state feedback so as to realize nonlinear control of rectifying side constant current and inverter side constant extinction angle. According to the nonlinear control method in the invention, accurate control of the rectifying side constant currentand the inverter side constant extinction angle and dynamic response characteristics of a polar layer and a valve layer in the direct current control system are obviously improved.