82 results about "Nonlinear networks" patented technology

Methods and systems for detecting defects on a specimen are provided. One system includes a generative model. The generative model includes a non-linear network configured for mapping blocks of pixels of an input feature map volume into labels. The labels are indicative of one or more defect-related characteristics of the blocks. The system inputs a single test image into the generative model, which determines features of blocks of pixels in the single test image and determines labels for the blocks based on the mapping. The system detects defects on the specimen based on the determined labels.

The invention discloses a fault-tolerant control method for a networked control system with time-varying delay. In view of conditions of parameter perturbation, time-varying delay, external disturbance and random failure happening to an actuator, a discrete time closed-loop nonlinear networked control system model is firstly built, a Lyapunov-Krasovskii function with the delay information is then built, a Lyapunov stability theory and a linear matrix inequality analysis method are used, sufficient condition for asymptotic stability of the nonlinear networked control system and existence of an H-infinity fault-tolerant controller are obtained, a Matlab LMI toolbox is used for solution, and a gain matrix for a non-fragile fault-tolerant controller is: K=K<->P; and conditions for optimizing the minimum disturbance rejection rate gamma is given, and the controller gain matrix K<*> optimized under the minimum disturbance rejection rate gammamin= square root of e is acquired. The condition of time-varying delay existing in the system is considered, the time-varying delay is analyzed and processed based on a free-weighting matrix method, and the conservation is reduced.

The invention provides a fault detection method of a nonlinear network control system based on an event triggering mechanism, and relates to the technical field of network system fault detection. Themethod comprises the following steps: firstly, establishing a T-S fuzzy model of the nonlinear network control system, setting an event triggering condition, establishing a fuzzy fault detection filter model, establishing a fault weighting system, and then establishing a fault detection system model; selecting an appropriate residual evaluation function and a detection threshold according to the fault detection system model, and detecting whether a fault of the nonlinear network control system occurs; and finally, further designing a parameter matrix and an event triggering matrix of a fault detection filter according to the stability of the fault detection system and sufficient conditions of existence of the fault detection filter. By adoption of the fault detection method of the nonlinear network control system based on the event triggering mechanism provided by the invention, the robustness to external disturbance and communication delay is greatly improved, and the limited networkresources and computing resources can be saved by the application of the event triggering mechanism.

The invention discloses a nonlinear networked control system non-fragile H-infinity fault tolerance control method. Firstly a closed-loop nonlinear networked control system model is established for considering the situation of parameter perturbation, time delay and packet loss of the nonlinear networked control system and random fault of an actuator, and then a Lyapunov function including packet loss information is constructed. The sufficient conditions of nonlinear networked control system stochastic stability and existence of an H-infinity fault tolerance controller are obtained by using the theory of Lyapunov stability and a linear matrix inequality analysis method. Solving is performed by using a Matlab LMI tool box, a non-fragile fault tolerance controller gain matrix K=YP<-1><011> is given, the conditions for optimization of the minimum disturbance suppression ratio gamma are given, and the optimized controller gain matrix K* under the minimum disturbance suppression ratio gamma<min> (the square root of e) is acquired. The situation of the random fault of the actuator is considered, and the probability of the random fault meets BerRoulli distribution so as to have more practical meaning.

The present invention discloses a non-fragile dissipative filtering method of a nonlinear networked control system. The method comprises the steps of firstly establishing a nonlinear networked filtering error system model on the conditions of considering the time delay and the packet loss of the nonlinear networked control system and the perturbation of the filter parameters, then constructing a Lyapunov function, and then utilizing a Lyapunov stability theory and a linear matrix inequality analysis method to obtain the sufficient conditions of the mean square exponential stability of a nonlinear networked filtering error system and the existence of a non-fragile dissipative filter, utilizing a Matlab LMI tool kit to solve, and definding a non-fragile dissipative filter parameter matrix. The method of the present invention considers the random time delay and the pocket loss situations between the sensors and the filters, is suitable for the general dissipative filtering including the H-infinite filtering, and enables the conservatism of the non-fragile dissipative filter design to be reduced. Moreover, a non-modeling state of the system is considered when a full-order filter is designed, thereby being able to reduce the calculation burdens and the design cost.

A method for designing feedforward and feedback controllers for integration of stochastic sources and loads into a nonlinear networked AC/DC microgrid system is provided. A reduced order model for general networked AC/DC microgrid systems is suitable for HSSPFC control design. A simple feedforward steady state solution is utilized for the feedforward controls block. Feedback control laws are provided for the energy storage systems. A HSSPFC controller design is implemented that incorporates energy storage systems that provides static and dynamic stability conditions for both the DC random stochastic input side and the AC random stochastic load side. Transient performance was investigated for the feedforward/feedback control case. Numerical simulations were performed and provided power and energy storage profile requirements for the networked AC/DC microgrid system overall performance. The HSSPFC design can be implemented in the Matlab/Simulink environment that is compatible with real time simulation/controllers.

The invention discloses an active disturbance rejection control method of a spacecraft considering network transmission and actuator saturation. The method includes the steps that firstly, a proper transition process is arranged for a desired attitude of a system by designing a tracking differentiator, and meanwhile a differential signal of an expected value is obtained to prepare for subsequent controller design; and then a nonlinear sampling extended state observer is designed by using an attitude angle measurement signal output from a network protocol, real-time estimation of a state in a spacecraft system and nonlinear uncertain items formed by coupling, external interference and so on is carried out, and an estimated value of the nonlinear uncertain items is compensated to an error feedback control rate containing an anti-saturation compensator. The active disturbance rejection control method of the spacecraft considering network transmission and actuator saturation can not only avoid the adverse effect of nonlinear factors such as internal and external interference on the system, but also ensure that an actuator can precisely control the spacecraft attitude within the saturation range, and provide guarantee for successful completion of space operation tasks. The active disturbance rejection control method of the spacecraft considering network transmission and actuator saturation has good control effect, and can be widely used in other nonlinear networked control systems.

The invention relates to a robust fault detection method of a discrete time networked system with random packet losses. The object is a time-varying system with unknown disturbance and a nonlinear term, parameters of the system vary along with time and meet norm bounded conditions, the nonlinear term meets fan-shaped domain bounded conditions, and the unknown input is any energy bounded signal with unknown statistical property. Through state augmentation, an original fault detection problem can be converted into a robust H8 filtering problem, namely, a robust fault detection filter is designed, and for any allowed time-varying parameter and external disturbance and all possible data losses, the difference between output of the fault detection filter and a fault signal is made to be as small as possible in the sense of H8. The Lyapunov function method is used for proving that a whole fault detection system is stable in asymptotic mean square and meets the corresponding H8 disturbance attenuation requirements; a full robust fault detection filter existing condition meeting requirements and a fault detection filter parameter design method are provided in an LMI mode.

The invention discloses an event trigger control method of a saturated nonlinear networked industrial control system. The method comprises the following steps: establishing a saturated nonlinear networked industrial control system state space model, designing a saturated nonlinear networked industrial control system state feedback controller based on event triggering, establishing a closed-loop system state space model, analyzing the stability and the passivity of a closed-loop system, and solving a feedback controller of the saturated nonlinear networked industrial control system based on event triggering. By designing the state feedback controller based on event triggering, the problem that the system cannot continuously, effectively and safely operate due to the fact that the influenceof state saturation, actuator saturation and external interference on the control performance of the industrial control system is rarely considered at the same time in an existing control method of the industrial control system is solved. The method can be used for effective control and safe operation of a complex networked nonlinear industrial control system.

The invention provides a network intrusion detection method and system based on data enhancement and BiLSTM (Bidirectional Long Short Term Memory), and solves the problem that the existing network intrusion detection method is low in intrusion detection and recognition accuracy of minority class attack samples, firstly, network intrusion detection flow data is collected, then, preliminary feature extraction is performed, a training data set is formed, attack type data samples with small data volume are confirmed and then data enhancement is carried out, then a BiLSTM neural network model is constructed and iterative learning training is carried out, the model automatically extracts feature information at a higher level, high-dimensional nonlinear network traffic features can be better processed, manual limitation caused by the fact that traditional shallow machine learning depends on manual feature extraction is overcome, the problem that class distribution in a training data set is unbalanced can be solved through data enhancement operation, and the recognition accuracy of minority class attack samples is improved on the premise that the model keeps a high overall detection rate and a low false alarm rate.

The invention discloses an efficient Doherty power amplifier, which comprises a driving power amplifier and a final power amplifier; the driving power amplifier has a balanced amplifying structure composed of a power dividing bridge 1, a driving power amplifier 2, a nonlinear network 1, a combiner 1 and a load; and the final power amplifier has a Doherty amplifying structure composed of a power dividing bridge 2, a nonlinear network 2, a carrier power amplifier, a peak power amplifier and a combiner 2. The driving power amplifier 1 and the driving amplifier 2 both work in type B bias, so the linearity and effectiveness of driving power amplifier can be improved. The carrier power amplifier works in type B bias; and the peak power amplifier works in type C bias and utilizes asymmetrical Doherty structure, so the efficiency of final power amplifier in reduction of 8dB is greatly improved. The efficient Doherty power amplifier has the advantages that the driving power amplifier and the final power amplifier are both efficient in working states, which can efficiently amplify the peak-to-average ratio. Meanwhile, uniformity of batch production of such efficient Doherty power amplifier can be greatly improved by adding the nonlinear matching network in the separating end of the power distributing bridge in the Doherty structure.

The invention discloses a robustness fault-tolerant control method of a networked control system based on an interval type 2 T-S model. Aiming a type of non-linear network control system with uncertain parameters, the method considers network induced delay. When an executor fails, a suitable fonctionelle is constructed based on the interval type 2 T-S model, so as to introduce a free weight matrixto represent relationships of all items in the formula, and a state feedback controller that enables the system to stabilize gradually and meets performance indexes is designed by using inequality scaling, matrix decomposition technique, mutual convex combination, improved Jensen's inequality and parallel distribution compensation technique. The interval type 2 T-S fuzzy model can approximate tothe global non-linear system by using any accuracy of the local linear system, makes up the type 1 fuzzy set in terms of processing of uncertain information, and makes calculation less complex compared with the type 2 fuzzy set.

Radio Frequency (RF) signal network measurement data of a device under test are acquired by exciting the device using a modulated RF excitation signal, while measuring RF signal data at the signal ports of the device, measuring bias signal data, and processing the RF signal data and the bias signal data, providing the RF signal network measurement data of the device. By acquiring bias signal data, in particular by measuring variations in the bias signals, a more accurate and reliable characterization of the non-linear behavior of the device under test can be provided. A Non-linear Network Measurement System (NNMS) is arranged for acquiring the RF and biasing signal data and characterizing the non-linear signal behavior of a device under test.

The invention discloses a state estimation method for a non-linear networked system under a spoofing attack. The method comprises the following steps of: establishing a non-linear system model and a system state estimator model, introducing an event triggering mechanism, and establishing a network attack model based on the influence of the spoofing attack on network transmission data; designing a nonlinear system state estimator model under the spoofing attack and event triggering mechanism; utilizing a Lyapunov stability theory to obtain a sufficiency condition for ensuring the mean square stability of a system index; and finally, solving a linear matrix inequality to obtain event triggering parameters and state estimator gain. According to the method of the invention, the bandwidth can be effectively saved, the network load is reduced, the communication capability of a transmission channel is improved, the network bandwidth resource can be efficiently saved, and the network load is reduced; meanwhile, the event triggering mechanism and a quantization mechanism are introduced, so that the burden of network transmission can be effectively reduced.

The invention discloses a self-bias power management integrated circuit (PMIC) chip power supply which is composed of an inductor L of a power circuit of a switch-type inductive DC-DC or AC-DC converter, a master power switch S of the power circuit of the switch-type inductive DC-DC or AC-DC converter and a nonlinear network, wherein the master power switch S and the inductor L are combined to provide an efficient current source for the nonlinear network; the nonlinear network controls the on and off of the master power switch; and the nonlinear network is a network capable of controlling to switch between a through branch and a charging branch, and the nonlinear network controls the efficient current source to charge a chip power supply bypass capacitor. The PMIC chip power supply is built by aiming at the switch-type inductive DC-DC or AC-DC converter and making full use of inductor working properties, simplifies the existing two groups of circuits to build a PMIC chip power supply scheme, and enables the chip power supply to be built rapidly and reliably.

The invention discloses a machine learning based screw-type material distributor controller which comprises a signal acquisition module, a processing module, a neural network module, an iterative learning module, a storage module, a first connecting array, a second connecting array and an output module. An adopted dynamic recurrent Elman neural network maps the material level of a blanking bin, falling difference in the air, blanking rate, material density and the spiral blade diameter, thread pitch and maximum screw rod rotation speed of a helical conveyor into material aerial amounts, the iterative learning module during off-line training adjusts weights according to a gradient descent method, and the processing module conducts advanced closing control on the helical conveyor through theoutput module according to prediction values of the aerial amounts in the online blanking control process. The controller adopts a nonlinear network to model the blanking process, the trained networkcan accurately predict the aerial amounts in different blanking states, accordingly direct and accurate blanking can be achieved, the machine learning based screw-type material distributor controlleris suitable for small-batch production, and the blanking efficiency is improved due to the fact that a screw can keep high operating speed.

The invention, which belongs to the technical field of operation control of the power system, relates to a linearized power flow calculation method for a radial distribution network. On the basis of the traditional linear equation without considering the network loss and the interphase mutual impedance, linear expansion of the nonlinear network loss item is included by the equation and a linearized three-phase branch power flow equation is established; with active power loads and reactive power loads of all nodes, similar values of active powers and reactive powers of all branches and all nodevoltages are calculated to obtain a power flow calculation result of the radial distribution network. According to the invention, compared with the linearized branch flow equation without consideringthe network loss, the provided equation enables the computing precision to be improved substantially; an approximate solution of the power flow can be obtained directly without iterative solution, sothat computing is done quickly. The linearized power flow calculation method is suitable for the scene with strict performance requirements on the real-time online analysis of the distribution network.

The invention discloses a method for determining an inertial navigation algorithm framework based on a nonlinear network. Non-mechanical layout is reflected in that a conventional navigation calculation manner of calculating position and speed by utilizing Newtonian mechanics equation is abandoned; and acknowledge is mainly reflected in that a navigation calculation model is automatically established based on tools such as a neural network aiming at different inertial devices. The method has the following steps: 1, the precision is higher, and since a nonlinear network model of acknowledge inertial navigation is established by analysis and digging of mass data, the model is customized, and compared with an original mechanical layout universal model, the model can be better adapted to inertial devices, thus having higher precision; and 2, the capability is promoted, a neutral network training model processes in background, and along with the optimization of training data and training methods, the capability of the acknowledge inertial navigation system is constantly improved, and therefore, the system has online updating capability.

The invention relates to a distribution network sensitivity calculating method based on a linearized power flow, and belongs to the technical field of power system operation control. On the basis of the conventional linear equation without considering the network loss and interphase mutual impedance, a nonlinear network loss item is linearly expanded and contained in the equation, and a linearizedpower flow equation of a three-phase branch circuit is established. According to variations of active power and reactive power of all nodes, the variations of the active power and the reactive powerof the branch circuit and variation of node voltage are calculated so as to obtain the sensitivity of the branch circuit power and the node voltage relative to node injection. Compared with the powerflow equation without considering the network loss, the distribution network sensitivity calculating method provided by the invention has the advantages as follows: the calculation accuracy is improved and the calculation is rapid; the distribution network sensitivity calculating method is applicable to real-time online analysis of a distribution network and other scenes.

The invention discloses a networked spacecraft attitude control method based on a hybrid forced observer, and belongs to the field of spacecraft attitude control. The method comprises the steps: firstly, starting from a networked spacecraft attitude kinetic equation containing random spoofing attacks, establishing an expansion system of the networked spacecraft attitude kinetic equation; designinga static event triggering mechanism; secondly, designing a hybrid forced observer to observe unknown nonlinear terms and system states in a spacecraft attitude system by using sensing signals containing event triggering and network spoofing attacks; and finally, designing a composite controller based on the output value of the observer, so that adverse effects of network spoofing attacks and disturbance inside and outside a spacecraft networked system on the system are avoided, the robustness of the system is improved, the transmission of the data volume measured by a sensor is reduced, and aguarantee is provided for smooth completion of a space operation task. The method has a good control effect on a spacecraft networked attitude control system considering network spoofing attacks, andcan be widely applied to other networked control systems containing multiple nonlinearity.

The invention discloses a weak nonlinear network control method based on a direct inverse control algorithm. The weak nonlinear network control method comprises the following steps of (1) calculating the spatial position of a network signal; (2) compensating an error of the network signal by means of the least-square interpolation estimation algorithm, replacing a current network signal detection result with an interpolation estimation result, obtaining a stability coefficient of the network signal by means of operation, and obtaining a weak nonlinear network signal control basis; (3) building a weak nonlinear control inverse model; (4) building a weak non-linear network signal control model. According to the weak nonlinear network control method, control over a weak nonlinear control system is conducted by means of the direct inverse control algorithm, and the signal to noise ratio of the network signal can be increased; meanwhile stability of the network signal is improved, and therefore the requirement for control over the network signal is met.

The invention belongs to the technical field of cross-domain recommendation algorithms, and particularly relates to a cross-domain recommendation method based on a stacked auto-encoder. Aiming at theproblem of data sparsity existing in pure cross-domain recommendation, the invention provides the cross-domain recommendation method based on the stacked auto-encoder, which can improve the score prediction accuracy and the classification accuracy of recommendation. According to the invention, the two models of the cross-domain stacked auto-encoder based on the user and the cross-domain stacked auto-encoder based on the project are learned at the same time, the learning results are compared, the optimal recommendation result is selected, and therefore the score prediction accuracy and the classification accuracy of recommendation are improved. According to the invention, cross-domain information is introduced into the automatic encoder so as to understand deeper nonlinear network structures of users and commodities. According to the invention, the sparsity problem is effectively solved by expanding the target domain user vector and combining deep learning, and the method is superior toother models in the aspects of score prediction and Topn recommendation.

The invention mainly relates to the technical field of image processing. The invention provides an image compressed sensing method based on a self-adaptive nonlinear network. The method comprises thesteps: partitioning an original image to obtain at least one original image block, measuring the original image block through a preset sampling rate based on a convolutional neural measurement network, obtaining a measurement value of the convolutional neural measurement network; reconstructing the measurement value through a full connection layer; calculating to obtain an approximate solution ofthe original image block according to the measured value; learning a residual error between the original image block and an approximate solution of the original image block through an SRCNN network model; and obtaining a reconstructed image according to the residual error, enabling the calculation of image reconstruction to be simpler, shortening the time of image reconstruction, and further improving the quality of the reconstructed image through bilinear interpolation and extended convolution by utilizing the characteristic that the measured value extracted by the measurement network also retains the spatial information of the image.

The invention discloses a Gaussian filtering method based on a nonlinear network system under a non-ideal condition, and relates to a Gaussian filtering method of a nonlinear network system. The objective of the invention is to solve the problems that related noise, one-step random delay measurement and data packet loss which may occur in a nonlinear network system are not considered in an existing method, and the problem that delay measurement based on model linear approximation may lead to filter estimation precision reduction and even divergence. The Gaussian filtering method based on the nonlinear network system under the non-ideal condition comprises the following steps: 1, establishing a system model and a sensor measurement model; 2, giving a hypothesis and a lemma; 3, designing a Gaussian filter based on the step 2; and 4, approximating the Gaussian weighted integral in the step 3 based on a third-order sphere diameter volume rule to obtain a numerical form of a designed filter. The method can be applied to the technical field of spacecraft and aircraft navigation.