Neural networks for control

End milling operations are essential to machine complex-contoured parts which are widely used in automobile and aerospace industries. Controlling the cutting force is important in these operations. But this is a difficult task for a number of reasons. A neural network control scheme is developed in this study to control the average resultant cutting force. The control scheme involves a neural identifier and a neural controller. The neural identifier is trained to represent the process accurately, and the neural controller is trained to give an input to the process which will yield the desired output. Recurrent neural networks are employed for both the neural identifier and neural controller, and training is accomplished using a recursive least squares algorithm. Results validating the methodology are presented for one-dimensional milling.

This paper describes the design and implementation of a neural network controller capable of printing designs and drawings on paper using a Cartesian Plotter. The Cartesian plotter uses two dc motors with incremental encoders for motion along x and y-axis while third dc motor along z-axis controls the pen assembly for drawing the text and graphics data. The motion commands for the motors are generated with the help of a simple feed forward neural network controller trained with back propagation algorithm. The nine inputs, two output neural controller employs four neurons in the hidden layer with tangent sigmoid being activation function for hidden neurons while linear function is used for output neurons. The proposed neural controller can drive the pen assembly in one of the eight possible directions and thus produces similar results to the famous Bresenham's algorithm. Simulations and practical results validate the neural controller for use with Cartesian plotters.

The ramifications of incorporating neural networks into two model-based control architectures, namely internal model control (IMC) and model predictive control (MPC), are considered. The development of a neural network analog to the conventional IMC design is described, the controller behavior is discussed, and control architectures necessary to improve controller performance are presented. The performance of the neural network controller under less restrictive assumptions is examined, and a neural network analog to the conventional MPC design is developed and tested. An IMC-type neural network controller in which the process model replaced by a neural network gives very good performance, even when only partial state data are available, also gives excellent performance. These results indicate that neural networks can learn accurate models and give good nonlinear control when model equations are not known. Suggestions for improving performance are presented. >

The present paper investigates an advanced control system of the wind energy conversion using an Artificial Neural Network (ANN) controller, which is developed in two modes: the offline mode, required for testing different sets of neural network parameters to find the optimal neural network controller (structure, activation function, and training algorithm), and the online mode where the optimal ANN controller is used in the five-phase PMSG based wind turbine system. The required data to generate the ANN model is obtained from a conventional PI controller. The controller inputs are the error between reference and actual speed (e), and the error change (de = e(k)-e(k-1)). As for the ANN output, it is the decreasing or increasing electromagnetic torque (dTe*). The studied neural network controller has been tested and validated using different wind speed values. Results and analysis are presented, and the contribution has been demonstrated.

This article analyses the dynamic behaviour of a beam structure containing multiple transverse cracks using neural network controller. The first three natural frequencies and mode shapes have been calculated using theoretical, finite-element, and experimental analysis for the cracked and un-cracked beam. Comparisons of the results among theoretical, finite-element, and experimental analysis have also been presented. The calculated vibration signatures were used to train the feed-forward multi-layered neural network controller with back-propagation technique for the prediction of cracks. Relative crack locations and relative crack depths are the output of the neural controller. Results obtained from the various analyses are validated using the developed experimental set-up. Results from neural controller have been presented for comparison with the output from theoretical, finite-element, and experimental analysis. From the evaluation of the performance of the neural network controller it is observed that the developed method can be used as a crack diagnostic tool in the domain of dynamically vibrating structures.

Objective: BLDC motor specifically used for high speed applications and needs sophisticated speed control. Conventional controls techniques require more physical involvement for assisting BLDC motor to operate with precise speed value. However, soft tuning based controller may give comparatively better results for that and the same has been elaborated in present paper using neural controller of speed as per requirements.the main objective of this study may be summarized as (i) MATLAB simulink model for BLDC motor incorporated with soft tuned controller and inverter simultaneously, (ii) development of neural controller for BLDC motor control, and (iii) validation of results developed through proposed controller with the available results using PI, PID and fuzzy logic controller. Methods: The speed obtained through simulink model with input supply voltage variations are used as a reference signal for the training and testing purpose of neural controller. With the help of these reference signals, neural control executes and gives optimum values of external parameters such as speed and electromagnetic torque. Result: The study outcome provides dynamic execution of proposed controllerbased on neural network. The proposed controller’s results are validated with the reference results obtained mathematically for the system. The controller efficiency and sensitivity has been checked by MATLAB-Simulink software. Novelty: comparative study for mathematically obtained reference values and neural controller based results are proposed in this paper. The results are validated with the help of Matlab produces tables and figures. Keywords: BLDC Motor; speed control; neural network controller

DC-DC converters can be found in almost every power electronics device. To improve the efficiency and controller response of a DC-DC converter to dynamical system changes, neural network has been chosen as an alternative to classic methods. However, no prior work has been done in the neural network approach for control of a PSFB (phase-shifted full-bridge) converter yet. In this research, a multi-layer feedforward neural network controller is proposed. The neural network based controller has the advantage of adaptive learning ability, and can work under the situation when the input voltage and load current fluctuate. Levenberg-Marquardt back-propagation training algorithm is used in computer simulation. The neural controller is then implemented on hardware using a DSP (digital signal processor). Satisfactory experimental results are obtained.

Adaptive growing-and-pruning neural network control for a linear piezoelectric ceramic motor

Model predictive control has been widely used in the industry. This can control the multivariable system with constraints on input and output variables but it needs online computation solver, and creates the non-convex solution in nonlinear plant due to the parameter uncertainties. The online computational problem and non-convex solution of the model predictive control are achieved via neural network model predictive control. The paper explores the speed control of permanent magnet synchronous motor (PMSM) by using neural network model predictive control (NNMPC) technique. The multi-layer artificial neural network is used to identify the dynamics of PMSM. The set point speed tracking control of PMSM is identified by using neural network model predictive control strategy. By using the set of input and output data obtained from the system, the multi input-output feed-forward neural network model is created. Levenberg-Marquardt algorithm is used to train the process models of the PMSM. That provides future plant output for control optimization of the predictive control. The overall system is developed and tested in the MATLAB/Simulink. To evaluate the efficiency of the controller proposed, it is compared with a constrained model predictive controller through the studies of simulation. The overshoot and settling time of the speed response of the PMSM are measured and analyzed for NNMPC and constrained MPC.

In this paper, adaptive neural network (NN) control is investigated for a class of multiinput and multioutput (MIMO) nonlinear systems with unknown bounded disturbances in discrete-time domain. The MIMO system under study consists of several subsystems with each subsystem in strict feedback form. The inputs of the MIMO system are in triangular form. First, through a coordinate transformation, the MIMO system is transformed into a sequential decrease cascade form (SDCF). Then, by using high-order neural networks (HONN) as emulators of the desired controls, an effective neural network control scheme with adaptation laws is developed. Through embedded backstepping, stability of the closed-loop system is proved based on Lyapunov synthesis. The output tracking errors are guaranteed to converge to a residue whose size is adjustable. Simulation results show the effectiveness of the proposed control scheme.

This benchmark suite presents a detailed description of a series of closed-loop control systems with artificial neural network controllers. In many applications, feed-forward neural networks are heavily involved in the implementation of controllers by learning and representing control laws through several methods such as model predictive control (MPC) and reinforcement learning (RL). The type of networks that we consider in this manuscript are feed-forward neural networks consisting of multiple hidden layers with ReLU activation functions and a linear activation function in the output layer. While neural network con- trollers have been able to achieve desirable performance in many contexts, they also present a unique challenge in that it is difficult to provide any guarantees about the correctness of their behavior or reason about the stability a system that employs their use. Thus, from a controls perspective, it is necessary to verify them in conjunction with their corresponding plants in closed-loop. While there have been a handful of works proposed towards the verification of closed-loop systems with feed-forward neural network controllers, this area still lacks attention and a unified set of benchmark examples on which verification techniques can be evaluated and compared. Thus, to this end, we present a range of closed-loop control systems ranging from two to six state variables, and a range of controllers with sizes in the range of eleven neurons to a few hundred neurons in more complex systems.

An adaptive sliding mode controller using radial basis function (RBF) network is proposed to approximate the unknown system dynamics for cantilever beam. Neural network controller is designed to approximate the unknown system model. In the presence of unknown model uncertainties and external disturbances, sliding mode controller is employed to compensate for such system nonlinearities and improve the tracking performance. On-line neural network (NN) weight tuning algorithms are designed based on Lyapunov stability theory, which can guarantee bounded tracking errors as well as bounded NN weights. Numerical simulation for cantilever beam is investigated to verify the effectiveness of the proposed adaptive neural control scheme and demonstrate the satisfactory vibration suppression performance.

An adaptive neural network control scheme is developed for a class of nonlinear systems in the strict-feedback form. Compared with the existing approaches, the main advantage is that the developed scheme can be implemented by utilizing only one neural network approximator. Thus, the designed controller structure is simplified. In addition, less neural network can reduce the running cost in practical application. The developed neural network control scheme can achieve that all the signals of the closed-loop system are uniformly bounded and the tracking errors converge to an arbitrary small neighborhood around zero by selecting suitably design parameters.

A neural network semi-active control strategy is proposed to control MR dampers in this paper. Firstly, the neural network is applied to calculate the unknown seismic wave via the much stronger generalized capacity of BP error retroactive transmission network, on the basis of which the "switch-consumption" semi-active control strategy is dependent upon the calculated value of the seismic wave. Then, the neural network controller's training course is accomplished with the help of this controlling strategy and calculating network. Lastly, the structure is manipulated intellectually with the controller. With the application of the above-mentioned controlling method is figured out an simulating result of roof MR intelligent isolation system, demonstrating: With the neural network semi-active control strategy, the system can prevent the seismic whiplash effect of the top machinery building of ship lift structure and can restrain the displacement between layers and the moment at the bottom of the pillar are obviously restrained

Recurrent fuzzy-neural approach for nonlinear control using dynamic structure learning scheme