Neuro Fuzzy Controller

Introduction

The introduction of fuzzy logic in the neural network area was, in its early stages. It is concerned with the development of multi-input and output neuron models. In contrast to the previous model, they developed a binary neuron model to provide median values ​​within the range. A link between neural physiology and fuzzy set theory has emerged since then to analyze sensory systems using concepts such as abstract language, abstract entropy, and abstract automata.

Fuzzy Logic may be a specific area of concentration within the learning of AI and is predicated on worth of that knowledge which is neither definitely true nor false. The knowledge which humans use in their everyday lives to base intuitive decisions and apply general rules of thumb can and will be applied to those control situations which demand them.

A neural network is a system made up of many simple processing tools that work in a similar way whose function is determined by network structure, communication capacity, and processing performed on the processing elements or neurons.

Fuzzy systems and neural networks have attracted the interest of researchers in various sciences and engineering fields. Number and variety of non-neural and neural logic use networks have been expanding, ranging from consumer products and industrial processes control in medical tools, information systems and decision-making.

The combination of neural networks and fuzzy logic offers the opportunity to solve tuning problems and complexity of intricate logic design. The resulting network will be more transparent and can be easily detected by means of unintelligible mind control rules or semantics. This is the new method that incorporates the well-established benefits of both methods and avoids the failures of both.

Characteristics:

Compared to a normal neural network, the weights of the connection and distribution and the activation functions of the fuzzy neural networks are very different. Although there are many different ways to model an abstract neural network, most of them agree on certain aspects such as the following:

• The neuro-fuzzy system based on the fuzzy system is trained in a learning process driven by data based on neural network theory. This heuristic considers only local knowledge in order to cause local changes in an important fuzzy system.
• It can be represented as a set of fuzzy rules at any point in the learning process, i.e., before, during and after.
• So the system may be activated with or without prior knowledge in terms of fuzzy rules.
• The learning process is mandatory to validate the semantic structures of the neuro fuzzy system.
• The neuro-fuzzy system is almost identical to the anonymous n-dimensional function represented in part by training models.
• Fuzzy rules can therefore be interpreted as vague examples of training data.

Architecture:

This structure can easily be understood as a “neural-like” architecture. At the same time, it can easily be interpreted as an Fuzzy logic control.

Figure 1: Architecture of fuzzy controller from neural networks point of view

1. The structures given in Figure 1 of the fuzzy logic control resemble a neural feedforward network.
2. The X-, R-, and C modules can be viewed as neurons in a neural layered network and µ- and ν-units as a flexible network.
3. The X module layer can be easily identified as input layer of multi-input neural network while the C module layer can be seen as the output layer.
4. The R-module layer serves as a hidden or central layer that creates internal network representation.
5. The fact that one µ-module can be connected to more than one R-module is equivalent to a connection to a neural network that shares same weight. This is very important in maintaining the integrity of the Fuzzy control structure.

Advantages

• No need of mathematical model
• Expert knowledge and experience used
• Can control non-linear plant
• Can control fast process

Disadvantages

• Cannot show stability of controlled system
• Even if operator exist, human knowledge is incomplete.
• No standard and systematic method for transferring human knowledge
• Computing time could be long

Applications:

1. NFS in Medical system

Currently diseases in India have emerged as leading killer in the urban and rural areas of the country. There will be high number if diseases are found at early age. Proper diagnosis of disease will reduce risk of death from various diseases. For past ten years neuro applications that have been misunderstood in the medical system are receiving lot of attention and that is why so much important research has been done. NFS is used for various common diagnostic disorders such as psychiatric disorders, heart disease, breast cancer, Alzheimer’s, thyroid disease, etc.

2. NFS in Economic system

An economic system can be defined as an organization in which a person, country or region makes, distributes, consumes, buys or sells goods and services. NFS can be used in various sectors of the economic system such as government economy, stock market, toll collection, gas condensate, gas consumption, power consumption, power forecasting, pricing forecast, supply chain management etc. made, including stock market.

3. NFS in traffic control

Recently a number of researchers have paid close attention to this field. Road management is a system used to describe how councils and highway authorities regulate use of road network to improve road safety and efficiency. Network traffic control is the process of managing, prioritizing, controlling or reducing network traffic to reduce congestion and delays.

4. NFS in image processing and feature extraction.

Image analysis is the process of extracting data from images using image processing techniques. Visual imaging fields and medical images for tasks require complex calculations to extract quantitative information and use pattern recognition, digital geometry, and signal processing. Other applications related to image processing include emotion recognition, image steganalysis, sound image processing, facial recognition and photo compression.

References

[1] http://www.scholarpedia.org/article/Fuzzy_neural_network

[2] https://www.omega.co.uk/technical-learning/pid-fuzzy-logic-adaptive-control.html

[3] Kar, Samarjit; Das, Sujit; Ghosh, Pijush Kanti (2014). Applications of neuro fuzzy systems: A brief review and future outline. Applied Soft Computing

[4] Gurpreet S. Sandhu and Kuldip S. Rattan Department of Electrical Engineering Wright State University Dayton, Ohio. Design of a Neuro Fuzzy Controller

[5] F.Gomide, A.Rocha, P.Albertos. UNICAMP. FEE/DCA. Campinas-SP. Brazil. Universidad Politecnica de Valencia, Spain. Neurofuzzy Controllers