A new online fuzzy control scheme based on reference model

In this study, using a model reference adaptation law, a stable fuzzy neural control system is developed. Despite the advantages of Model reference control design technique, which is mainly its power to exactly set trajectories of the system under control, this method is designed for linear system. In this study using fuzzy neural systems, a stable model reference controller for nonlinear systems is developed, Lyapunov method is used to guarantee the stability of fuzzy neural training algorithm and model following of the system under control.

Fuzzy logic-based control has emerged as a promising approach for complex and/or ill-defined process control. In this paper, a self-learning fuzzy controller with neural plant estimator is designed for the snack food frying control and the specific objectives are as follows: 1) to find the control variables affecting on product quality based on the statistical results of experimental data; 2) to employ the neural estimator for the prediction of real plant output related to time lag; 3) to construct the adaptive-network-based fuzzy inference system for the fuzzy inference rule extraction and the membership function tuning; and 4) to evaluate the designed controller performance by simulation. >

The antilock braking system (ABS) is designed to optimize braking effectiveness and maintain steerability; however, the ABS performance will be degraded in the case of severe road conditions. In this study, a self-learning fuzzy sliding-mode control (SLFSMC) design method is proposed for ABS. The SLFSMC ABS will modulate the brake torque for optimum braking. The SLFSMC system is comprised of a fuzzy controller and a robust controller. The fuzzy controller is designed to mimic an ideal controller and the robust controller is designed to compensate for the approximation error between the ideal controller and the fuzzy controller. The tuning algorithms of the controller are derived in the Lyapunov sense; thus, the stability of the system can be guaranteed. Also, the derivation of the proposed SLFSMC ABS does not need to use a vehicle-braking model. Simulations are performed to demonstrate the effectiveness of the proposed SLFSMC ABS in adapting to changes for various road conditions.

EAF steelmaking is a complex physical and chemical reactions，has Highly nonlinearity, time variability and strong coupling. As the electrode's contact with the charging circuit, charging the collapse of the gasification furnace components and liquid metal boiling and other factors, often result in dramatic fluctuations in arc current. At this point we need to quickly adjust the location of electrodes, so that arc current stable base on this within a certain range. This article in the fuzzy control and neural network’s foundation,will study a kind based on three-phase consciousness electric arc furnace predictive control new method, Using the fuzzy neural network control adjusting electrode, further enhancing the effectiveness of electric arc furnace’synthesis movement.

This paper presents the design and application of fuzzy neural network control system for the rotary cement kiln system. Due to the dynamic characteristics and reaction process parameters are with features of large inertia, pure hysteresis, nonlinearity and strong coupling, the fuzzy neural network controller combining both the advantages of neural network and fuzzy control was applied. The fuzzy neural network is an adaptive control process whose parameters can be adjusted by learning algorithms automatically which is aimed to eliminate the shortcomings of conventional control methods. The main control system structure includes mainly the pressure control loop, the burning zone control loop and the back-end of kiln temperature control loop. Simulation results show the effectiveness of the control scheme with satisfied dynamic performances on response time and overshoot.

A hybrid fuzzy neural architecture is proposed. The fuzzy neural system is a feedforward network that combines the basic notions of neural networks and fuzzy logic into a common structure. A back-propagation algorithm is used to train the FNS to perform the desired nonlinear mappings. Four simulation examples are presented where the fuzzy neural system is employed as a nonlinear identifier. Finally, comparisons between fuzzy neural systems, backpropagation neural networks and other fuzzy systems are given and discussed.

In this paper the identification and control of dynamic plants using type-2 TSK fuzzy neural system (FNS) is considered. The systems constructed on the base of type-1 fuzzy systems cannot directly handle the uncertainties associated with information or data in the knowledge base of the process. One possible way to alleviate the problem is to resort to the use of type-2 fuzzy systems. In this paper, a type-2 TSK fuzzy neural system (FNS), is proposed and its gradient learning algorithm is derived. Its performance for identification and control of time-varying plants is evaluated and compared with other approaches seen in the literature; the time-varying nature of the plants being handled as uncertainties in the plant coefficients which can be described by type-2 fuzzy sets.

Parameter estimation of fuzzy neural network controller based on a modified differential evolution

Investors react to news, particularly to earnings and dividend announcements released by respective firms, and consequently stock prices move. Thus, news has an impact on stock prices. However, the price adjustment process is a complex one. While neural fuzzy systems have advantages over statistical methods in modeling and predicting complex relationships generally, not many neural fuzzy systems share the same level of competence and capabilities. In this study, we evaluate the effectiveness of four neural fuzzy systems - feed forward neural network (FFNN), adaptive neuro fuzzy inference system (ANFIS), radial basis function network (RBFN) and rough set based pseudo outer product rule (RSPOP), respectively - in predicting the impact of news on stock price movements. We found that rough set based pseudo outer product rule (RSPOP) is the most effective system in the study undertaken, and is a candidate for further evaluation as a financial intelligence system.

This paper presents a self-learning fuzzy neural control scheme for a complicated object-an arc welding process. In this scheme, an analytical formula of fuzzy controller is used for adaptively regulating the fuzzy rules, and a neural network model is used for real-time learning dynamics of the uncertain process. The scheme has been used in control of welding pool dynamics of the arc welding process, and experiment results show the effectiveness of the control scheme.

This paper proposes neural network controllers using a reference model and investigates their characteristics. Since the neural network compensates the incorrectness of the plant such as nonlinearity and uncertainty against the reference model, a conventional controller can be designed based on the known dynamics of the reference model. The neural network controller can be classified into three types in terms of the connection between the neural network controller and the conventional controller. In the D-type controller, the output from the conventional controller inputs to the neural network and the reference model in parallel. In the Y-type controller, the output from the conventional controller inputs to the neural network, the plant and the reference model in parallel. The neural network of both controllers learns the combination of the inverse characteristics of the plant and the forward characteristics of the reference model. In the I-type controller, the output from the conventional controller inputs to the reference model whose output inputs to the neural network. The neural network leams the inverse characteristics of the plant. Simulation results using nonlinear plants show the achievement of the neural controllers and confirm their effectiveness for controlling the nonlinear plants.

A neural fuzzy system can learn an agent profile of a user when it samples user question-answer data. A fuzzy system uses if-then rules to store and compress the agent's knowledge of the user's likes and dislikes. A neural system uses training data to form and tune the rules. The profile is a preference map or a bumpy utility surface defined over the space of search objects. Rules define fuzzy patches that cover the surface bumps as learning unfolds and as the fuzzy agent system gives a finer approximation of the profile. The agent system searches for preferred objects with the learned profile and with a new fuzzy measure of similarity. The appendix derives the supervised learning law that tunes this matching measure with fresh sample data. We test the fuzzy agent profile system on object spaces of flowers and sunsets and test the fuzzy agent matching system on an object space of sunset images. Rule explosion and data acquisition impose fundamental limits on the system designs.

This chapter shows how neural fuzzy systems can improve motion estimation and compensation for video compression. Motion estimation and compensation are key parts of video compression. They help remove temporal redundancies in images. Most motion estimation algorithms neglect the strong-temporal correlations within the motion field. The search windows stay the same through the image sequences and the estimation needs heavy computation. We use an unsupervised neural vector quantizer system can use the temporal correlation of the motion field to estimate the motion vectors. First-order and second-order statistics of the motion vectors give ellipsoidal search windows. This algorithm reduces the search area and entropy and gives clustered motion fields. Motion-compensated video coding further assumes that each block of pixels moves with uniform translational motion. This often does not hold and can produce block artifacts. We use a supervised neural fuzzy system to compensate for the overlapped block motion. This fuzzy system uses the motion vectors of neighboring blocks to map the prior frame’s pixel values to the current pixel value. The neural fuzzy system used 196 rules that came from the prior decoded frame. The fuzzy system learns and updates its rules as it decodes the image. The fuzzy system also improved the compensation accuracy. The appendix derives both the fuzzy system and the supervised neural-like learning laws that tune its parameters.

Automated mechanical clutch is a critical component in vehicle powertrain, whose operations have great influence on fuel economy, comfort, and drivability. However, the control of clutch is a challenging problem due to the nonlinearities and uncertainties, which has attracted extensive attention. This paper proposed a model reference self-learning fuzzy control method with considering clutch wear. Fuzzy control is a simple but effective method to deal with the nonlinear system, but the determined fuzzy rules cannot be suitable for system variations due to clutch wear. Therefore, a self-learning algorithm based on referent model is introduced into fuzzy controller. Then, simulations are carried out in MATLAB/SIMULINK, and the results show that the method solves the problem of clutch wear effectively compared with the fuzzy control without self-learning ability. Finally, a test bench is designed and experiments are carried out to verify the validity of the proposed method.

While artificial neural networks and fuzzy systems have both been used as universal approximators, the two approaches have different advantages. For example, neural networks are good at classification and learning, while fuzzy systems can perform inference. To take advantage of such complementary strengths, various hybrid neural-fuzzy systems have been devised. The research reported here involves a new combination of neural and fuzzy systems developed for the protein folding problem, that is, how to estimate the number of topological hydrophobic contacts in the (unknown) most stable conformation of a given sequence of monomer residues. Fuzzy meta-rules are used to generate a series of neural networks for longer and longer input monomer sequences. >