SCIENCE TRIBUNE Thursday, October 24, 2002, Chandigarh, India

What is neuro-fuzzy logic?
Surjit Singh Bhatti
T is common now-a-days to come across electronic gadgets marketed by multinational companies that claim the use of "fuzzy logic" control systems. Nissan, for instance, has fixed fuzzy anti-lock brakes in their vehicles. Similarly, Hitachi and Matsushita air-conditioners, Canon copying machines, Toshiba and Sony micro-wave ovens, Samsung washing machines, among others, are examples of consumer products that use the fuzzy control devices.

Key to safeguarding secrets
DVANCES in devices that emit the smallest possible amount of light may portend an era of guaranteed confidentiality in Digital Age communication, which until now has had about as much protection from prying eyes as a message on a post card.


  • Helmet for directions

  • Artificial rain experiment

  • Contamination research




What is neuro-fuzzy logic?
Surjit Singh Bhatti

IT is common now-a-days to come across electronic gadgets marketed by multinational companies that claim the use of "fuzzy logic" control systems. Nissan, for instance, has fixed fuzzy anti-lock brakes in their vehicles. Similarly, Hitachi and Matsushita air-conditioners, Canon copying machines, Toshiba and Sony micro-wave ovens, Samsung washing machines, among others, are examples of consumer products that use the fuzzy control devices. Besides, automation is being achieved in factories and process industries using sophisticated fuzzy controls which are inexpensive and easier to maintain compared to the conventional "digital logic" control systems.

Fuzzy logic models itself on the pattern of human reasoning in its use of approximate information and uncertainty to generate decisions. It was designed (during late 1980s and early 1990s) to mathematically represent vagueness and develop tools for dealing with imprecision inherent in several problems. Normally, in digital computers one uses the "binary" logic where the digital signal has two discrete levels : low (logic zero) or high (logic one); nothing in-between. Fuzzy systems use soft linguistic variables (e.g. hot, tall, slow, light, heavy, dry, small, positive, ...etc.) and a range of their weightage (or truth) values, called membership functions, in the interval (0, 1), enabling the new computers to make human-like decisions. Since human beings tend to use words rather than numbers to describe behaviour patterns, fuzzy controls avoid the conventional rigidity of computers and allow them to use parameters based on "common sense."

Fuzzy logic application to a problem involves three steps: converting crisp (numerical) values to a set of fuzzy values, an inference system (based on fuzzy if-then rules) and de-fuzzification. In the first step, live inputs (such as temperature, distance, pressure, speed etc.) generate a real action in the form of pulse width measurement driving a motor or a voltage driving a motor (or a relay). This is done with the help of the special functions, called membership functions, which are the equivalents of adverbs and adjectives in speech; such as very, slightly, extremely, somewhat and so on. Next, the membership functions are put in the form of "if-then" statements and a set of rules is defined for the problem under investigation. These rules are of the form:

"IF A is low and B is high THEN C is medium."

Here A, B are input variables (known data) and C is an output variable (data value to be computed). The adjectives "low" in relation to A, "high" in relation to B and "medium" in relation to C are the membership functions in this case.

The IF is an antecedent, usually a sensor reading while THEN is a consequent, a command, in control applications. Each antecedent can lead to several premises and several consequences. This is not the case with binary logic where only one rule operates at a time. In Fuzzy logic more than one rule may operate at the same time but with varied strengths. This set of rules (with different weightages) leads to a crisp control action through the process of defuzzification. If one considers the case of a plane approaching a landing site, the inputs are its speed (A) and distance (B) while the output is the controlling power (C) required for accurate landing. The following table shows the possible 4x4 fuzzy C matrix with A shown horizontally and B vertically.

For a given speed and a given distance the values of input parameters are defined and the value of the output parameter, the braking power (required for accurate landing) has to be found. This is done by deciding if the speed is very slow, slow, fast or very fast and if the distance is very close, close, far or very far. The intersection of these vague linguistic parameters decides the output or power needed for correct control in a general way. For more accurate determination of the output, one has to assign numerical membership values to slow, fast, very slow and very fast speeds and, likewise, to close, far, very close and very far distances. It is clear from the above that fuzzy reasoning is neither exact nor in-exact. It is exact to a certain degree depending upon the choice of mathematical model and the membership functions. However, fuzzy control is more akin to human control than the purely yes-no, pass-fail, true-false, good-bad or win-lose type of binary logic.

In conventional electronic control systems, the output of the process under control is converted into an electrical signal output which is compared to a reference input signal. The difference or error signal actuates the controller to generate a control action signal. When the control signal is proportional to the error signal at a given moment, the output can take on any value between zero and one (fully off and fully on). If the control signal varies as the cumulative value of the error signals up to that moment, the control action is integral. It is also possible to have a control signal that depends upon the rate of change of error — giving a derivative controller. Generally, all the three types of actions (Proportional P, Integral I, and Derivative D) are present simultaneously in industrial controllers. The problem with PID controllers is that they follow binary logic. Whenever, fuzzy logic is made applicable in these PID control systems, the results are better as now an experienced human operator-like fuzzy rule-based system takes over the control from the water-tight binary logic.

Fuzzy logic finds applications in medical diagnostics where imprecision and uncertainty generally creep in despite best efforts. Past history of a patient may not be complete or may be given incorrectly. Mistakes may be made in physical examination and crucial symptoms may be overlooked inadvertently. Sometimes interpretation of investigations may not be right. Fuzzy set frame work has been utilised in several different approaches to "model" the diagnostic process. The information gathered by the different tests represents a fuzzy relation between the symptoms and the diseases. In this way, the physician is unlikely to go much astray although the diagnosis may not be completely accurate. This is a big gain as diagnosis "going astray" is much more dangerous for the patient than the diagnosis being slightly less accurate.

The writer is Professor, Applied Physics, Guru Nanak Dev University.

Neural networks

Speed (A)

Distance (B) Very Slow Slow Fast Very Fast
Very close Light Heavy Very Heavy Very Heavy
Close  Light Light Heavy Very Heavy
Far Light Very Light  Light Heavy
Very Far  Very Light  Very Light  Light Light

IN living organisms, specialised cells, called neurons, form a complex network which receives, processes and transmits information from one part of the nervous system to another. The centre of this network is in the brain which stores as well as analyses the information. Based on this analysis, the nervous system controls the various organs of the body. Neurons receiving stimulus produce electrical pulses, independent of the intensity of the stimulus, that propagate as in a cable and activate other neural chains, leading ultimately to an action potential. Electronic neural networks are inspired by their biological counterparts.

Both fuzzy logic and neural network concepts were developed during the last few years independently to understand human behaviour patterns, specially the thinking processes in relation to problem solving. While fuzzy logic uses approximate human reasoning in knowledge-based systems, the neural networks aim at pattern recognition, optimisation and decision making. A combination of these two technological innovations delivers the best results. This has led to a new science called neuro-fuzzy logic in which the explicit knowledge representation of fuzzy logic is augmented by the learning power of simulated neural networks. The computer results are thus much closer to human perception than ever before.



Key to safeguarding secrets

ADVANCES in devices that emit the smallest possible amount of light may portend an era of guaranteed confidentiality in Digital Age communication, which until now has had about as much protection from prying eyes as a message on a post card.

Experiments in emitting one photon — or packet — of light at a time point to ways to thwart even the most determined hackers’ attempts to eavesdrop on sensitive, private or other information not intended for public consumption, scientists said.

Communication signals carried by a single photon, as opposed to two or more, traveling through fiber-optic cables could provide an impregnable barrier against intruders bent on taking a look at wha they are not meant to see, they said.

Because ordinary light sources cannot generate the extraordinarily dim, single-photon pulses reliably, researchers have set their sights on other alternatives, including quantum dots — electron clusters so tiny, some 5,000 of them could stretch across a grain of sand.

These molecule-sized "boxes" that trap and release electrons — negatively charged elementary particles — could serve as light-emitting diodes, or LEDs, that efficiently turn an electrical field into photons.

Working with quantum dots, a group of Stanford University researchers has taken an important step toward an ideal single-photon source, one that would produce exactly one particle of light in a "pure" state, said team leader Charles Santori.

"Past reports claiming to demonstrate single-photon sources have only demonstrated control over the number of photons emitted at a time," Santori says. "In our report, we demonstrate a device for which consecutively emitted photons are usually in the same quantum state." The feat carries implications for the atom-scale field of quantum information.




Helmet for directions

Hoping to recapture those 2 to 4 seconds it takes to refocus on the road after a rider glances at the dashboard, the prototype Blue Eye helmet from DesignWorks/ USA and BMW features the world’s first motorcycle head-up display.

The 320-by 240-pixel color LCD is positioned 2 inches from your eye, close enough so it’s out of focus. But the info on the display is focused at infinity, so it remains clear as you look down the road. No production plans yet. For more

Artificial rain experiment

A cloud-seeding experiment, undertaken for the first time in Visakhapatnam, the water-starved port city of Andhra Pradesh, has become an instant success as the city experienced moderate rain hours after the endeavour.

Buoyed by the success, noted environmentalist T Shivaji Rao, who undertook the mission, said experiments would continue for some more days by blowing silver iodide fumes onto the cold clouds and common salt on the warm clouds to precipitate the rain.

Dr Rao, who is also the director of the Centre for Environmental Studies under the Gandhi Institute of Technology and Management (Gitam), told UNI that similar experiments were conducted in Hyderabad and drought-prone Anantapur district in 1995.

He claimed that it would cost only 75 paise to generate 1000 litres of rain water from the clouds. But unfortunately, the governments — both at the Centre and the states — had been neglecting the well-proven 40-year-old technology which had been widely used in several developed and developing countries, including neighbouring Pakistan and China besides Russia. UNI

Contamination research

Poisonous mercury is on dinner plates everywhere — in sea bass served in fancy restaurants, in tuna casserole ladled out at home. Most of the time, there is so little that it goes unnoticed. But that doesn’t mean it’s harmless. Eat enough and it can make you sick.

Too much mercury damages the nervous system, especially the brain. Too much in pregnant and breast-feeding women, or those who may become pregnant, can hurt their babies adversely affecting children’s intelligence, coordination and memory. Children under 7 are vulnerable too, because their young brains are still forming.

How much is too much? And are adults at risk, as well?

Rising public concern about those questions, which have been in the background for years, is prompting public health officials to look seriously at mercury and at its effects.

State and federal officials disagree over what constitutes a safe exposure level. AP





Across :

1. The air sacs contained in the lungs.

6. Symbol for Mercury.

8. Symbol for Barium.

9. Huge sheet of ice formed from compressed snow gliding down mountain slopes.

12. A linearly homogeneous production function having a constant elasticity of input substitution.

13. Trunk apart from head and limbs.

14. Lines drawn on maps to connect places of equal rainfall.

17. Naturally occurring oxides of Iron.

18. …….meter is used to measure the flow of liquids.

20. Symbol for Germanium.

21. …..plane is horizontal stabilizing surface of an aeroplane.

22. Equipment to drill an oil well.

24. Short for ultra-violet radiation.

25. A scheme of IDBI to encourage commercially proven advanced technology.

26. Abbreviation for SI unit of pressure.

27. A unit of conductance.

28. An important protein of muscles belonging to globulin class.


1. This number is reciprocal of dispersive power of a substance.

2. A regular network of geometrical points about which the atoms or molecules move in a crystal.

3. Early stage of a bird’s ovum.

4. Resinous substance secreted by some insects.

5. Huge mass of ice floating in the sea.

6. Symbol for Helium.

7. Derived SI unit of absorbed dose of ionizing radiation.

10. The curves between pressure and volume at constant temperature.

11. Lining of metal let into an orifice to guard against wearing by friction.

15. This apparatus is used to estimate carbon dioxide, oxygen and carbon monoxide in exhaust gases.

16. A non inflammable mono atomic gas.

19. Part of body serving some special function.

23. Indian institute conducting courses in study in plastics (abbr.)

Solution to last week’s