245 results about "Neural network controller" patented technology

Adaptive controller apparatus of a plant may be implemented. The controller may comprise an encoder block and a control block. The encoder may utilize basis function kernel expansion technique to encode an arbitrary combination of inputs into spike output. The controller may comprise spiking neuron network operable according to reinforcement learning process. The network may receive the encoder output via a plurality of plastic connections. The process may be configured to adaptively modify connection weights in order to maximize process performance, associated with a target outcome. The relevant features of the input may be identified and used for enabling the controlled plant to achieve the target outcome. The stochasticity of the learning process may be modulated. Stochasticity may be increased during initial stage of learning in order to encourage exploration. During subsequent controller operation, stochasticity may be reduced to reduce energy use by the controller.

The invention relates to a multipurpose intelligent control method for an electronic expansion valve of a heat pump system of a refrigeration air-condition, belonging to the technical field of the independent heat and cold energy supply of buildings. The method comprises the steps of(1) determining two or more controlled parameters affected by the electronic expansion valve,(2) determining a set value for each controlled parameter and carrying out detection in real time, (3)comparing the detection value and the set value of each controlled parameter, calculating deviation and deviation change rate, then carrying out indistinct processing to generate the control quantity of respective electronic expansion valve,(4) obtaining the control quantity of each electronic expansion valve by being calculated by the step (3) to be taken as the input layer of a neural network controller and obtaining the right value of each controlled parameter by being trained by a single neuron neural network and (5) obtaining the feedback control quantity of the electronic expansion valve according to each controlled parameter value detected in real time. Due to the adoption of the method, operating parameters can be intelligently adjusted, so that intelligent heat and cold supply and stable and efficient operation under a variety of climatic conditions are realized.

A family of novel multi-layer discrete-time neural net controllers is presented for the control of an multi-input multi-output (MIMO) dynamical system. No learning phase is needed. The structure of the neural net (NN) controller is derived using a filtered error / passivity approach. For guaranteed stability, the upper bound on the constant learning rate parameter for the delta rule employed in standard back propagation is shown to decrease with the number of hidden-layer neurons so that learning must slow down. This major drawback is shown to be easily overcome by using a projection algorithm in each layer. The notion of persistency of excitation for multilayer NN is defined and explored. New on-line improved tuning algorithms for discrete-time systems are derived, which are similar to e-modification for the case of continuous-time systems, that include a modification to the learning rate parameter plus a correction term. These algorithms guarantee tracking as well as bounded NN weights. An extension of these novel weight tuning updates to NN with an arbitrary number of hidden layers is discussed. The notions of discrete-time passive NN, dissipative NN, and robust NN are introduced.

A rotary-table servo system neural network control method comprises (1) building a mechanical dynamic model of a permanent magnet synchronous motor rotary-table servo system, and initializing the system state, sampling time and relative control parameters; (2) according to the differential mean value theorem, enabling a non-linear input dead zone in the system to linearly approximate to a simple time-varying system, avoiding complex calculation of dead-zone inverse compensation, and finally inferring a rotary-table servo system model provided with an unknown dead zone; (3) at each sampling moment, calculating and controlling a system tracking error, a fast terminal sliding mode surface and a first-order derivative of the system; (4) based on the rotary-table servo system model provided with the unknown dead zone, selecting a neural network approaching unknown trend, designing an adaptive robust finite-time neural network controller according to the system tracking error, the fast terminal sliding mode surface and the first-order derivative of the system, and updating a neural network weight matrix; and (5) entering the next sampling moment, and repetitively executing the steps from (3) to (5).

The invention discloses a robust neural network control system for a micro-electro-mechanical system (MEMS) gyroscope based on sliding mode compensation and a control method of the control system. The control system comprises a given trajectory generation module, a sliding mode surface definition module, a neural network controller, a weight adaptive mechanism module, a sliding mode compensator, an MEMS gyroscope system, a proportional-differential control module, a first adder and a second adder. The control method of the control system comprises the following steps of: establishing an MEMS gyroscope kinetic model based on a sliding mode surface, designing a controller structure, and designing an updating algorithm of a radial basis function (RBF) network weight, so that the trajectory of the MEMS gyroscope is tacked. By the control method, the influence of the unknown dynamic characteristic of the MEMS gyroscope and noise interference can be compensated on line, the vibration trajectory of the MEMS gyroscope completely follows a reference trajectory, and the anti-interference robustness and reliability of the system are improved; the updating algorithm of the network weight is designed on the basis of a Lyapunov stability theory, so that the stability of a closed-loop system is ensured; and a powerful basis is provided for expanding the application range of the MEMS gyroscope.

The invention relates to a data drive control method for minimum energy consumption of a refrigerating system on basis of SPSA. The method includes the steps of adjusting frequency of a compressor to enable chilled water supply water temperature to be constant according to changes of a system load so that refrigerating capacity can be matched with a thermal load; obtaining a relation curve between the system load and the minimum stable superheat degree of an evaporator; establishing an online neural network identification model of the system; calculating the system load according to changes of refrigerating capacity of the refrigerating system of an air conditioner under dynamic regulation of the compressor, obtaining the minimum stable superheat degree corresponding to the system load according to the relation curve between the system load and the minimum stable superheat degree of the evaporator, and using the minimum stable superheat degree as a set value of the superheat degree of the evaporator; establishing a neural network controller; completing control over the superheat degree of the evaporator through an expansion valve control loop. The method is easy to calculate and implement, the number of parameters is small, and the control effect is good.

The invention relates to an ultrasonic wave motor transient characteristic testing device and a control system of the device. The device comprises a base and an ultrasonic wave motor arranged on the base, wherein an output shaft on one side of the ultrasonic wave motor is connected with an optical-electricity encoder, an output shaft on the other side of the ultrasonic wave motor is connected with a flywheel inertia load, an output shaft of the flywheel inertia load is connected with a torque sensor through a coupler, the signal output end of the optical-electricity encoder and the signal output end of the torque sensor are respectively connected to the control system, and the control system is composed of an RNNI and an RNNC. The RNNI completes the identification of the input characteristics and the output characteristics of the ultrasonic wave motor under the different control variables and different flywheel inertia loads, and the RNNC achieves the speed / position / torque control output of the ultrasonic wave motor according to the identified result so as to determine the control dynamic characteristic under the different loads and the different control variables. The device and the control system are high in testing accuracy, simple and compact in structure and good in using effect.

The invention particularly discloses a mountain road sharp turn section real-time vehicle speed early warning method. The mountain road sharp turn section real-time vehicle speed early warning method comprises steps as below: step one: arranging basic information collection devices on vehicles and sharp turn roads, collecting information of roads and vehicles in real time; step two: constructing a calculation controller of road attachment coefficients and rolling resistance coefficients based on a BP neural network in a computer; step three: inputting each parameter collected in step one into the BP neural network controller for processing, outputting road attachment coefficients and rolling resistance coefficients of the vehicles at a curve position of the road; step four: calculating distances between the vehicles; step five: calculating the safety distances between the vehicles in the computer; step six: comparing the safety distance of the vehicles and the distances between the vehicles, then carrying out alarming for vehicles exceeding a limit speed.

A greenhouse intelligent control method is realized by taking a plant intelligent databank as a base to control the integral climate of the greenhouse, additionally by tracing the timely information of plant growing period to correct and enhance the databank and superposing chaos signals into the input signal for controlling the greenhouse climate and also adopting a nerve network controller for optimal regulation. The invention has the advantages that: 1. the method is green and environment-friendly, because the intelligent control technique is adopted to generate the green natural ecological environment with a chaos-phenomenon which further approaches the nature changing and is good for the optimal growth of plants; 2. the method can conserve energy, because the traditional greenhouse control system without chaos changing is changed so as to conserve a large quantity of energy used for keeping constant-temperature in the greenhouse. The invention puts forward a novel greenhouse intelligent control system simulating the natural environment on the basis of taking the greenhouse plant growth environment as a background, with excellent effect, practical application and popularization value.

The invention discloses a welding equipment network monitoring device and a control method thereof. A signal acquisition device acquires key characteristic operation parameter signals of field welding equipment and transmits the key characteristic operation parameter signals to a signal conditioning circuit; the signal conditioning circuit receives, denoises and zooms to acquire conditioned parameter signals and transmits the conditioned parameter signals to a micro controller; the micro controller operates the conditioned parameter signals to acquire actual working parameters of the field welding equipment; a network controller transmits welding process specification to the micro controller; when the actual working parameters cannot meet the welding process specification, the micro controller transmits alarm information to an LCD (Liquid Crystal Display); and the LCD displays abnormal working parameters to realize monitoring. According to the welding equipment network monitoring device and a control method thereof, the network monitoring of the field welding equipment is realized; and the device has simple design, is low in cost and has portability.

The invention discloses a mechanical arm general control method based on determined learning theory. The method comprises the following steps: establishing a mechanical arm dynamic model, establishing an expected period trajectory; establishing an adaptive RBF (radial basis function) neural network controller, adjusting a weight of the RBF neural network controller, thus conditions that a mechanical arm tracks the expected period trajectory and the RBF neural network locally approaches an unknown dynamic model in a mechanical arm closed-loop system; establishing a constant neural network; and using the constant RBF neural network to finish a control task. By using the mechanical arm general control method provided by the invention, the experience period trajectory of the mechanical arm closed-loop control system unknown dynamic along the mechanical arm can be accurately learned in a local region under a condition that a system parameter is completely unknown; the effective knowledge of the closed-loop system dynamics can be learned in a stable dynamic control process, and can be stored in a manner of constant RBF network weight; the effective knowledge can be successfully applied to the subsequently same or similar control task so as to improve the control performance of the control system and save energy.

The invention provides a Q-function self-adaptation dynamic planning method based on data. The Q-function self-adaptation dynamic planning method based on the data achieves an optimal control aim. The Q-function self-adaptation dynamic planning method based on the data mainly comprises the following steps: (1) initializing a stable control strategy, (2) initializing a weight value of an actuator and critic neural network through the existing control strategy, (3) according to the current control strategy and a system state of a current state, generating a control motion of a controlled system, exerting the control motion on a controlled member, observing a system state of a next time, (4) regulating the weight value of the actuator and critic neural network, (5) judging whether a current iteration cycle is finished or not, if the current iteration cycle is finished, entering step (6), if the current iteration cycle is not finished, returning to the step (3), (6) judging whether neural network weight values generated by the two closest iteration cycles are obviously changed, if the neural network weight values generated by the two closest iteration cycles are obviously changed, entering the step (2) through a newly generated actor and critic neural network, and if the neural network weight values generated by the two closest iteration cycles are not obviously changed, outputting a final actor and critic neural network.

A neural network control-based reload robot position controller comprises a neural network controller and a plurality of servo drivers. The neural network controller provides control information for the servo drivers. Each servo driver comprises a speed controller, a position controller and a current controller. The neural network controller adopts an Elman network in a dynamic recursive neural network structure to overcome the dynamic system identification problems, guarantee convergence rate and simplify learning rule method. An input layer is connected with the output of a sensor, and input information including position, speed and current signals is detected and controlled through the sensor. The Elman network improves local feedback and self-feedback elements in the self-learning manner, thus the robot control network is imparted a memory function to adapt to complicated dynamic environment.

The invention discloses a non-position sensing control method for an electric vehicle drive motor. The method comprises the steps of using a pulse voltage injection method for detecting an initial position of a drive motor rotor before an electric vehicle is started; and in order to conduct accurate detection, using a rotary high-frequency voltage injection method at a low speed section for real-time detection of the position of the drive motor rotor after the electrical vehicle is started; in order to avoid multiple salient pole effects of a motor, adding a double salient pole decoupling observer onto a structure of the high-frequency voltage injection method, at an intermediate speed, combining the high-frequency injection method and the counter electromotive force method and through a neural network controller to process and detect the position of the drive motor rotor, and at a high speed, using the counter electromotive force method to detect the position of the drive motor rotor; and meanwhile, adding a rotor position robust observer to solve the problem that the observer converges statuses of an opposite position to a rotor magnetic pole due to external interference at the high speed. According to the invention, the rotor position information of the electric vehicle drive motor can be accurately and effectively detected.

The invention discloses a control method for sewage processing process based on the self-organizing neural network, and belongs to the fields of water treatment and intelligent information control. The method mainly comprises adjustment for fuzzy rules by a self-organizing mechanism and self-adaptive learning control of T-S fuzzy neural network. The method comprises the steps that comparison is carried out on the basis of a T-S fuzzy neural network controller; self-organizing adjustment is carried out on the fuzzy mechanism; self-adaptive learning of the neural network is carried out; and the fuzzy rule m at the time k is obtained, and the sewage treatment process at the time k is controlled. The method can be used to adjust the internal structure of the controller in real time according to the environment state, and an object is controlled stably. The self-organizing mechanism is used to adjust the controller structure in real time so that the controller can satisfy environment requirements more effectively; the intelligent control method can be used to control the sewage treatment process stably, so that the quality of output water meet the discharge standard; and the defect that a controller of a fixed network structure is low in environment adaptability is overcome.

The invention relates to a mechanical shoulder joint position control method with dynamic friction compensation, which is realized through a global control unit and a local control unit. The global control unit is used for tracking the trajectory of a mechanical shoulder joint in a global large range, the trajectory tracking is realized through a PD (Proportional Differential) controller widely applied in the mechanical shoulder joint, and the input vector of the PD controller comprises the position error of the mechanical arm joint and the change rate of the position error; and the local control unit is used for completing dynamic friction compensation in a local small range, the dynamic friction compensation is realized through a five-layer autoregressive wavelet neutral network controller having an observation layer, and the input vector of the autoregressive wavelet neutral network controller comprises the expected position, the expected speed and the actual position of the mechanical shoulder joint. The actual speed of the mechanical shoulder joint required in the autoregressive wavelet neutral network controller can be calculated through the observation layer. The mechanical shoulder joint position control method provided by the invention can be realized by only installing one position sensor in the mechanical shoulder joint without installing a speed sensor.

An artificial multiped is constructed (either in simulation or embodied) in such a way that its natural body dynamics allow the lower part of each leg to swing naturally under the influence of gravity. The upper part of each leg is actively actuated in the sagittal plane. The necessary input to drive the above-mentioned actuators is derived from a neural network controller. The latter is arranged as two bi-directionally coupled chains of neural oscillators, the number of which equals twice that of the legs to be actuated. Parameter optimisation of the controllers is achieved by evolutionary computation in the form of a genetic algorithm.

The invention belongs to the field of distributed power generation and smart power grid control, relates to a neural network-based microgrid energy coordination central processor, and particularly provides a smart energy central controller which can utilize a BP (Back Propagation) neural network to control a microgrid comprising a wind power generation unit, a gas turbine generation unit, a PV (photovoltaic) generation unit, a load tap changing transformer, various loads and other units. In the implementing process, actual power of each micro-power supply, loaded actual power and system frequency are utilized as 9 input nodes of the neural network controller; and a network mapping relationship between the 9 input nodes and 11 output nodes is constructed by using a single hidden layer BP neural network containing 12 hidden layer nodes, so that energy balance management of a microgrid system is controlled, and self-adaptive connection between the microgrid and large grids is realized.

The invention discloses a performance limited flexible manipulator control method based on determining learning theory. Aiming at the indeterminacy of the dynamic model of the flexible manipulator, the method designs a tracking error to satisfy the limitation of constraint conditions, and forms an error controller. The method in the invention comprises the following steps: building a dynamic model of the flexible manipulator, building a system state observer, designing tracking error performance constraint conditions, designing a neural network controller based on the determining learning theory, and correcting the controller utilizing experiential knowledge. The method designed in the invention can realize the dynamic properties of rapid convergence and low overshoot, satisfy the limitation of set constraint conditions, avoid the on-line regulations to the neural network weights at the same time, and shorten the control time. In addition, the method can utilize the learned experiential knowledge to control the later same task rapidly.

The invention discloses an energy management strategy for an airplane adaptive power and thermal management system (APTMS), and belongs to the technical field of airplane integrated heat / energy. According to the invention, firstly, an APTMS energy optimization rule is obtained through off-line simulation by means of an instantaneous optimization energy management strategy and by combining a plurality of working conditions, then energy management rules are classified by means of fuzzy C-means clustering, and a part of the rules are extracted to serve as training samples of a nerve network. A BP nerve network controller obtained through training controls energy distribution of the system in dependence on the real-time working condition of the APTMS in order to achieve energy optimization management. The energy management strategy for the airplane adaptive power and thermal management system (APTMS) can guarantee fuel economy of the APTMS, and the energy management real-time performance is obviously improved.

The invention discloses an adaptive inverse control system for a chemical-mechanical polishing machine, which comprises a target pressure input device used for inputting target pressure of a polishing head chamber to a compound control device, a neural network identification device used for identifying a control model of a system based on the input pressure of the polishing head chamber input by the compound control device and the pressure output by the polishing head chamber, and the compound control device used for outputting the input pressure of the polishing head chamber based on the target pressure, the pressure output by the polishing head chamber and the control model. Each compound control unit comprises a neural network controller used for calculating neural network controlled quantity, a proportional integral controller used for generating proportional integral controlled quantity and a weighting device used for generating the input pressure of the polishing head chamber for the neural network controlled quantity and the proportional integral controlled quantity with the subsection variable-parameter control strategy. The adaptive inverse control system for the chemical-mechanical polishing machine is capable of reducing coupling among areas of the polishing head chamber through dynamic on-line decoupling of neurons.

The invention discloses a fractional order terminal sliding mode-based AFNN control method of an active power filter. The method comprises the steps of designing a mathematical model of an active filter, a fractional order-based nonsingular terminal sliding mode controller and a fractional order-based adaptive fuzzy neural network controller; and controlling the active power filter by using output of a fractional order-based nonsingular terminal sliding mode adaptive fuzzy neural network controller. According to the AFNN control method, the disadvantage that a nonsingular inversion terminal sliding mode control strategy needs accurate system information is overcome and the robustness is improved; good performance can still be kept when an external load changes; operation of the active power filter along a sliding mode track is ensured through designing the sliding mode controller; for the disadvantages of an inversion control law, an AFNN controller is adopted to approach a nonlinear part in the active power filter. A fractional order module is introduced into the sliding mode controller and the adaptive controller, so that an adjustable item is added by a fractional order in comparison with an integer order, and the overall performance of a system is improved.

The invention relates to a control method of a light-type direct-current transmission system converter of an offshore wind power station, belonging to the technical field of power transmission. In the invention, a PID (Piping and Instruments Diagram) neural network controller is designed based on a particle group optimizing method, the traditional PI (Piping and Instruments) regulator is substituted, step input is used to train a neural network, that is to say, multi-group particles are set to search in the neural network weight space of searching, and the position and speed of the particles are continuously updated according to an adaptive value function of the neural network to obtain the optimum weight of the neural network. The invention adopts the method of using the optimum weight value obtained by training in combination with the error forward broadcast of the neural network to substitute the traditional PI regulator to control the operation of the system, thereby reducing the parameter to be regulated and improving the transient response performance of the system; the neural network weight value is obtained by training a nonlinear model of a controlled system so as to approach to a true system. Only the forward broadcast process of the PID neural network is controlled in the operation process of the system. The method is relatively simple and is easy to achieve.

The invention discloses an adaptive fuzzy sliding mode RBF neural network control method for an active power filter. The control method is characterized by comprising the following steps of step 1, establishing an active power filter mathematic model; step 2, obtaining an adaptive fuzzy sliding mode RBF neural network controller based on fuzzy sliding mode design, including a fuzzy adaptive law and an RBF neural network adaptive law; and step 3, controlling the active power filter according to the adaptive fuzzy sliding mode RBF neural network controller. According to the control method, the command current can be tracked and compensated in real time; and in addition, the control method is high in reliability, high in parameter change robustness and high in stability.

The invention provides a neural network self-adaptive control method applied to a servo system, aiming to improve the control accuracy of the servo system. The neural network self-adaptive control method realizes nonlinear compensation and interference suppression to the servo system and improves the tracking accuracy of the servo system. The servo system mainly comprises a reference model, a self-adaptive controller and a neural network controller. In addition, based on a speed ring accurate reference model, a position ring controller is very simple in design so that the whole system is rather convenient in design, thereby being easy to carry out in engineering.

The invention provides a temperature control method and system. The system comprises a temperature sensor, a neural network controller and a temperature adjusting device. The temperature sensor is used for obtaining a current temperature value of the environment; the neural network controller is used for substituting the current temperature value of the environment and a target temperature value into a preset temperature model to generate a temperature control signal; and the temperature adjusting device is used for adjusting the current temperature value of the environment according to the temperature control signal to enable the current temperature value of the environment to be equal to the target temperature value. The environment temperature is controlled through the neural network controller capable of carrying out self learning, so that control precision of the temperature control system can be improved, adjusting time is reduced, and stability precision is improved; and throughthe self learning method, the temperature control system can carry out self learning on temperature variation rules of different environments, so that manual intervention is not necessarily need, anduniversality is improved.

The invention discloses a mode-based intelligent control method for a position-limited flexible joint robot. The method comprises the following steps that a kinetic model of the flexible joint robot and a universal model of a plurality of expected regression trajectories are established; constrained conditions of position limitation are given based on actual working conditions, corresponding position transfer functions are designed, and a constant-value neural network controller group satisfying transient performance requirements and based on the a reference pattern is designed according to adetermined learning theory; a dynamic pattern library of the expected regression trajectory model is established; a pattern recognition scheme and controller switching strategy based on dynamic patterns is designed. The mode-based intelligent control method for the position-limited flexible joint robot has the advantages that the flexible joint robot acquires and utilizes empirical knowledge fromcomplex work tasks when output variables meet the specified constrained conditions, thereby achieving the real-time monitoring and autonomous rapid identification of the dynamic pattern of the system,also ensuring that the smooth continuity of input signals is controlled during mode switching, and providing a guarantee for the stability of the control system.

The invention discloses a decoupling control method of a single leg joint of a hydraulic four-leg robot. The method comprises proportion integration differentiation (PID) neural network decoupling control, neural network model reference decoupling control and prediction control. The method comprises the following steps that a reference model under the condition that a system is free of coupling is set, and then a neural network controller is trained so that the output of the system can keep up with the output of the reference model; meanwhile, a neural network model is used for prediction, and the next step of output of the system is predicted according to current and previous input and output data of a controlled object; and finally, the weight of the neural network is rectified in an online mode according to the predicted output and the given reference output so that optimizing indexes of the decoupling controller of the neural network can reach the smallest value, and the purpose of decoupling control is achieved. By the adoption of the decoupling control method, the coupling influences among all the joints of the robot can be effectively reduced, and decoupling control over all the joints of the robot is achieved.

The invention discloses an adaptive neural network control method for an arc micro-electromechanical system, which comprises the steps of a, building a system model of the arc micro-electromechanicalsystem based on the Bernoulli beam; b, constructing an adaptive neural network controller used for suppressing chaotic oscillation of the arc micro-electromechanical system and guaranteeing state constraints of the system, wherein when the adaptive neural network controller is constructed, output constraints of the arc micro-electromechanical system are ensured not to be violated by using a symmetrical obstacle Lyapunov function, an unknown non-linear function is estimated with an arbitrary small error by adopting an RBF neural network with an adaptive law, an extension state tracking differentiator is introduced to process a problem that virtual control items in backstepping control need to be derived repeatedly, a state observer is designed to obtain unmeasured state information, the extension state tracking differentiator and the state observer are integrated in the backstepping framework. The adaptive neural network control method has the characteristics of convenient stability analysis and proving, low requirement for the modeling precision, low computation complexity, high operation speed, good operation stability of the system and high motion accuracy.