A novel SPSA kernel wavelet neural network for model-free PID controller

This paper proposes the model-free PI control design for direct current (DC) drives. A model-free PI control system structure is considered in the framework of reference tracking control using a first-order nonlinear dynamic system as a local approximation of the process model. The derivatives are estimated numerically using a Savitzky-Golay filter that solves the differentiation and also the smoothing. The model-free PI control system structure is applied to the speed control of a laboratory nonlinear DC drive as a representative mechatronics application. The experimental results for three different reference input shapes show the very good performance in model reference tracking, reference trajectory tracking and numerical differentiation.

This research mainly based on multilayer perceptron (MLP) neural networks technique of data mining to forecast the wheat crop yield at the district level. There are many statistical and simulation models available, but the proposed algorithm with new activation function provides promising results in a shorter time with more accuracy. Sigmoid and hyperbolic tangent activation functions are widely used in the neural network. The activation functions play an important role in the neural network learning algorithm. The main objective of the proposed work is to develop an amended MLP neural network with new activation function and revised random weights and bias values for crop yield estimation by using the different weather parameter datasets. MLP model has been tested by existing activation functions and newly created activation functions with different cases including weights and bias values. In this research study, we evaluate the result of different activation functions and recommend some new simple activation functions, named DharaSig, DharaSigm and SHBSig, to improve the performance of neural networks and accurate results. Also, three new activation functions created with little variations in the DharaSig function named DharaSig1, DharaSig2 and DharaSig3. In this research study, variable numbers of hidden layers are tested with the variable number of neurons per hidden layer for the agriculture dataset. Variable values of momentum, seed and learning rate are also used in this study. Experiments show that newly created activation functions provide better results compared to ‘sigmoid’ default neural network activation function for agriculture datasets.

Application of Neural Networks in Computer Security

Active deployment of machine learning systems sets a task of their protection against different types of attacks that threaten confidentiality, integrity and accessibility of both processed data and trained models. One of the promising ways for such protection is the development of privacy-preserving machine learning systems, that use homomorphic encryption schemes to protect data and models. However, such schemes can only process polynomial functions, which means that we need to construct polynomial approximations for nonlinear functions used in neural models. The goal of this paper is the construction of precise approximations of several widely used neural network activation functions while limiting the degree of approximation polynomials as well as the evaluation of the impact of the approximation precision on the resulting value of the whole neural network. In contrast to the previous publications, in the current paper we study and compare different ways for polynomial approximation construction, introduce precision metrics, present exact formulas for approximation polynomials as well as exact values of corresponding precisions. We compare our results with the previously published ones. Finally, for a simple convolutional network we experimentally evaluate the impact of the approximation precision on the bias of the output neuron values of the network from the original ones. Our results show that the best approximation for ReLU could be obtained with the numeric method, and for the sigmoid and hyperbolic tangent – with Chebyshev polynomials. At the same time, the best approximation among the three functions could be obtained for ReLU. The results could be used for the construction of polynomial approximations of activation functions in privacy-preserving machine learning systems.

Robust Control Theory Based Stability Certificates for Neural Network Approximated Nonlinear Model Predictive Control

Similar to many complex systems, the operation process of electric balanced forklifts has characteristics such as time-varying model parameters and nonlinearity. Establishing an accurate mathematical model becomes challenging, making it difficult to apply model-based control methods in engineering practice. Aiming at the longitudinal control system of electric forklifts containing external disturbances, this paper proposes an improved full-format dynamic linearization model-free adaptive PID control (iFFDL-MFA-PID) method. Firstly, the full-format dynamic linearization (FFDL) method is employed to transform the operating system of the electric balanced forklift into a virtual equivalent linear data model. Secondly, the nonlinear residual term and pseudo-gradient (PG) of the data model are estimated using the difference estimation algorithm and the optimal criterion function, respectively. Furthermore, in order to enhance the robustness of the system, the idea of intelligent PID (iPID) is introduced and the principle of equivalent feedback is utilized to derive the iFFDL-MFA-PID control scheme. The design process of this scheme only requires the use of the input and output data of the system, without relying on the mathematical model of the system. Finally, the iFFDL-MFA-PID method proposed in this paper is simulated and tested with the EFG-BC/320 counterbalanced forklift equipped in the Special Equipment Testing Center and compared with the model-free adaptive control method (FFDL-MFAC) and the PID control method. Simulation results show that the speed-tracking error of the electric forklift truck under the action of the iFFDL-MFA-PID algorithm is maintained within ±0.132 m/s throughout the process, achieving higher tracking accuracy and better robustness compared to the MFAC and PID methods.

Vibration control based on a virtual controlled object (VCO) is a simple and practical model-free active vibration suppression strategy. To utilize VCO-based model-free control, controller tuning plays the critical role. This study presents a computationally efficient tuning strategy for the VCO-based model-free active vibration controller. A state equation for model-free controller design is derived based on an actuator model and the VCO. The parameters of the VCO are determined so as to achieve the vibration suppression in the predetermined controlled frequency band. Then, we introduce the reference controlled object (RCO). The control performance of the VCO-controller is evaluated through the vibration control simulation for the RCO. Iteratively executing the vibration suppression simulation for the RCO, the suitable design parameters of the VCO-controller are searched on the basis of the simultaneous perturbation stochastic approximation (SPSA) algorithm. The nonlinear parametrization technique enhances the tuning efficiency. The validity of the tuning scheme is demonstrated by applying it to the design problem of the VCO-based model-free linear quadratic regulator (LQR). The numerical simulations verify the applicability of the VCO-LQR tuned by the proposed approach to wide variety of controlled objects other than the RCO.

There are a lot of methods for the stabilization of quadcopters and the newest are based on AI. A neural network is a simplified model that imitates the human brain's processes. In the research paper, we present a neural network control model for quadcopter stabilization. A single hidden layer network model was estimated to investigate the dynamics of the UAV. A control system with a classical PID controller was used to train the neural network model. This method is used for examining how the neural network imitates the stabilization of the quadcopter in real flight mode. The novelty of the work was to design of small size 3 layers NN model that runs in real-time in a quadcopter. The PID and machine learning controllers' operation results were compared to each other andshown in the experiment.

This paper considers the interconnection of a continuous time linear time-invariant system and a multivariable magnitude nonlinearity. A number of different quadratic constraints are established for the magnitude nonlinearity and then used to derive stability criteria based on quadratic and Lurie-type Lyapunov functions. The new stability criteria are cast as matrix inequalities and in some cases solved using semi-definite programming. Connections are made between the magnitude nonlinearity and neural network activation functions, such as the rectified linear unit (ReLU) and leaky ReLU, effectively allowing the stability criteria derived here to be used to analyse interconnections of dynamical systems and neural networks. Using the positive homogeneity property, shared by the (leaky) ReLU and magnitude functions, mild conditions are also established to show that the existence of a unique equilibrium point is sufficient for local and global stability to be equivalent. Finally, the new global stability criteria are tested on several numerical examples, including a Hopfield network with 100 states and neurons, and compare favourably with competing criteria from the literature.

The aim of this paper is to propose a neural network based reference modified PID control which has a simple duration design method for transient characteristics improvement of digitally controlled dc-dc converters. In renewable energy network systems, various types of dc-dc converters are widely used for power conversion and such converters require a superior control method for a stable operation. Especially, transient characteristics should be improved since they heavily affect the stability of the system. For such purposes, designing of conventional control methods becomes a difficult task since the optimization of control parameters needs complicated analysis and it is affected from variations of circuit components of converters. Therefore, simple and easy design of control is widely required for a stable operation of power converters. The neural network can provide a suitable control methodology for such situation since it treats the plant as a black box and it can realize a non-linear control based on training of the input-output relation without complicated modelling and analysis. On the other hand, the neural network based method has a disadvantage caused from the fact that the neural network is trained with data obtained in advance and an overcompensation phenomenon occurs in the transient response. In this paper, the neural network control is adopted to control the dc-dc converter in coordination with a conventional PID control. The neural network predicts the output voltage of the converter and the reference value in the PID control is modified with the predictions to reduce the error of the output voltage. To avoid overcompensation, a simple duration design for the neural network control is also provided to improve the transient response effectively. From prototype testing in simulation and experiment, it is revealed that the proposed method contributes to obtain a superior transient performance compared with the conventional PID control.

Fractional order PID, which involve integration and differentiation of non-integer order, is increasingly being used in the fields of control system, robotics, signal processing and circuit theory. Based on neural network, this paper introduce a new approach for design fractional order PID controller. A self-learning PID controller with five dimension parameters is realized by using parameter turning strategy of RBF neural network. Further apply the fractional order PID controller to a typical FOLPD process like pulp consistency control system. In particular, pulp consistency system is a typical FOLPD process. The simulation results show that this kind of controller has a better effect for pulp consistency system than traditional PID and PIDNN controller.

Considering the problem of the highly dependence on the accuracy model information of the industrial control process, an online PID parameter tuning method is proposed based on the simultaneous perturbation stochastic approximation (SPSA) algorithm and the traditional PID control only by the input and output data of the system. Interfering with each component of the PID controller parameters simultaneously, and the gradient information of the control performance loss function is estimated via the output value after disturbance by the global optimization method of stochastic approximation algorithm. The PID controller parameter is iterative updated according to the gradient estimation value. Finally, we applied this method to the control process of cement grinding particle size, and the simulation results by MATLAB shown it can lead a high control precision which proves the feasibility and effectiveness of this adaptive PID control method.

Fault Detection and Diagnosis in Deep-trough Hydroponics using Intelligent Computational Tools

In this paper, a high-performance optimal fractional emotional intelligent controller for an Automatic Voltage Regulator (AVR) in power system using Cuckoo optimization algorithm (COA) is proposed. AVR is the main controller within the excitation system that preserves the terminal voltage of a synchronous generator at a specified level. The proposed control strategy is based on brain emotional learning, which is a self-tuning controller so-called brain emotional learning based intelligent controller (BELBIC) and is based on sensory inputs and emotional cues. The major contribution of the paper is that to use the merits of fractional order PID (FOPID) controllers, a FOPID controller is employed to formulate stimulant input (SI) signal. This is a distinct advantage over published papers in the literature that a PID controller used to generate SI. Furthermore, another remarkable feature of the proposed approach is that it is a model-free controller. The proposed control strategy can be a promising controller in terms of simplicity of design, ease of implementation and less time-consuming. In addition, in order to enhance the performance of the proposed controller, its parameters are tuned by COA. In order to design BELBIC controller for AVR system a multi-objective optimization problem including overshoot, settling time, rise time and steady-state error is formulated. Simulation studies confirm that the proposed controller compared to classical PID and FOPID controllers introduced in the literature shows superior performance regarding model uncertainties. Having applied the proposed controller, the rise time and settling time are improved 47% and 57%, respectively.

Secondary path modeling is an important step in active noise control. Keeping the active noise control system stable, when the secondary path changes over time, is a widely studied problem. Using the simultaneous perturbation stochastic approximation algorithm, can update the control filter without secondary path modeling, avoiding the problem of increased noise caused by secondary path modeling. In this paper a variant of the simultaneous perturbation stochastic approximation algorithm is proposed, which incorporates historical coefficients of the control filter, to correct the direction of the gradient calculation. Based on an experimental setup with a realistic transfer path, this paper demonstrates that the proposed algorithm has superior performance in terms of convergence speed and noise reduction amount, than the general simultaneous perturbation stochastic approximation algorithm.