Fuzzy Logic, Fuzzy Sets, Fuzzy Rules, and Applications

TAKEAWAYS:

• Fuzzy logic extends classical logic by allowing partial truth values (between 0 and 1) to model uncertainty and vagueness in real-world data.
• Fuzzy sets represent elements with degrees of membership (e.g., "tall" = 0.7) instead of binary membership (0 or 1).
• Fuzzy inference systems use rules like "IF temperature IS high AND humidity IS low THEN fan_speed IS medium" to make decisions under uncertainty.
• Defuzzification converts fuzzy outputs into crisp actions (e.g., "turn fan speed to 60%").
• Applications include air conditioners (adjusting temperature based on fuzzy rules), washing machines (optimizing cycles), and medical diagnosis systems.
• Advantages over crisp systems: handles imprecise data, mimics human reasoning, and works well with sensor-based real-time decisions.

---

1. Introduction to Fuzzy Logic and Fuzzy Sets

Classical (crisp) logic uses binary values: true (1) or false (0). However, real-world data is often vague or imprecise. For example:

• "Is 170 cm tall?" → Crisp logic says no (if the threshold is 180 cm), but intuitively, 170 cm is somewhat tall.
• Fuzzy logic allows degrees of truth (e.g., 0.6 for "tall").

Fuzzy Sets vs. Crisp Sets

---

2. Membership Functions: How Fuzzy Logic Quantifies Vagueness

A membership function (μ) assigns a degree of membership to each input. Common shapes:

1. Triangular: Simple, defined by three points (e.g., "low," "medium," "high" speed).
2. Trapezoidal: Flat top for a range of values (e.g., "very hot" from 35°C to 45°C).
3. Gaussian: Smooth bell curve (e.g., "pressure is normal").

Example: Defining "Speed" for a Car Assume speed ranges from 0 to 120 km/h. Define fuzzy sets:

• Slow: Triangular (0, 0, 40)
• Medium: Triangular (30, 60, 90)
• Fast: Triangular (80, 120, 120)

For speed = 50 km/h:

• μ(Slow) = 0 (outside range)
• μ(Medium) = (60 - 50)/(60 - 30) = 0.67
• μ(Fast) = 0

---

---

3. Fuzzy Inference System (FIS): Rules and Reasoning

A fuzzy inference system mimics human decision-making using:

1. Fuzzification: Convert crisp inputs to fuzzy sets.
2. Rule Evaluation: Apply fuzzy rules (IF-THEN statements).
3. Aggregation: Combine rule outputs.
4. Defuzzification: Convert fuzzy output to a crisp action.

Example: Air Conditioner Control

Inputs:

• Temperature (T) ∈ [20°C, 40°C]
• Humidity (H) ∈ [30%, 90%]

Fuzzy Sets:

• T: {Cold, Warm, Hot}
• H: {Low, Medium, High}

Rules (5 examples):

1. IF T = Hot AND H = Low THEN Fan = High
2. IF T = Warm AND H = Medium THEN Fan = Medium
3. IF T = Cold AND H = High THEN Fan = Low
4. IF T = Hot AND H = High THEN Fan = Very High
5. IF T = Warm AND H = Low THEN Fan = Low

Worked Example:

• Crisp Inputs: T = 32°C, H = 40%
• Fuzzification:
  • μ(Hot) = 0.8 (from triangular function peaking at 35°C)
  • μ(Low) = 0.6 (from trapezoidal function for humidity ≤ 50%)
• Rule Firing:
  • Rule 1 fires with strength = min(0.8, 0.6) = 0.6
  • Rule 5 does not fire (H is not Low).
• Aggregation: Combine outputs (e.g., weighted average).
• Defuzzification: Centroid method → Fan speed = 75%.

---

4. Defuzzification Methods

Convert fuzzy outputs to crisp values. Common methods:

1. Centroid (COG): Calculate the center of gravity of the output.
   • Formula: $z^{\ast }=\frac{\sum \mu (z)\cdot z}{\sum \mu (z)}$
2. Max-Membership: Choose the z with the highest μ(z).
3. Weighted Average: Use rule strengths as weights.

Example (Centroid Method): Suppose the output fuzzy set for "Fan speed" has:

• μ(50%) = 0.3
• μ(75%) = 0.8
• μ(100%) = 0.1 Then: $z^{\ast }=\frac{(0.3\times 50)+(0.8\times 75)+(0.1\times 100)}{0.3+0.8+0.1}=70\%$

---

5. Fuzzy Expert Systems vs. Traditional Expert Systems

---

6. Applications of Fuzzy Expert Systems

In the Real World

1. Air Conditioners (e.g., Daikin, LG)

   • Idea Used: Fuzzy logic adjusts temperature and fan speed based on imprecise inputs like "room feels warm" (sensor data + user feedback).
   • How: Rules like "IF temperature > 28°C AND humidity > 60% THEN cooling_power = high."

2. Washing Machines (e.g., Samsung, LG)

   • Idea Used: Fuzzy control optimizes wash cycles based on dirty level (fuzzy sets: "light," "medium," "heavy") and fabric type.
   • How: "IF dirt = heavy AND fabric = cotton THEN water_level = high AND time = long."

3. Stock Market Prediction (e.g., Trading Bots)

   • Idea Used: Fuzzy rules analyze vague indicators like "market sentiment is bearish" (combining news, volume, and price trends).
   • How: "IF sentiment = negative AND volume = high THEN sell 30%."

4. Traffic Light Control (e.g., Smart Cities)

   • Idea Used: Adjusts signal timings based on fuzzy traffic density (e.g., "lane is congested" = μ = 0.7).
   • How: "IF north_south_traffic = high AND east_west_traffic = low THEN green_time = long (north-south)."

5. Medical Diagnosis (e.g., Fuzzy-Based AI Tools)

   • Idea Used: Handles uncertain symptoms (e.g., "fever is slightly high" = μ = 0.6).
   • How: "IF fever = medium AND headache = present THEN probability(dengue) = 0.7."

---

---

7. Designing a Fuzzy Expert System: Step-by-Step

1. Problem Identification: Define the system goal (e.g., "optimize AC cooling").
2. Input/Output Variables: List variables (e.g., temperature, humidity, fan speed).
3. Fuzzification: Define membership functions for each input/output.
4. Rule Base: Create IF-THEN rules (e.g., 10–20 rules for a simple system).
5. Inference Engine: Implement rule evaluation (e.g., min/max operators).
6. Defuzzification: Choose a method (centroid, max-membership).
7. Testing: Validate with real-world data (e.g., AC in a room).

Worked Example: Simple Loan Approval System Inputs:

• Credit Score (CS): {Poor, Fair, Good, Excellent}
• Income (I): {Low, Medium, High}

Output:

• Loan Amount (LA): {Small, Medium, Large}

Rules:

1. IF CS = Excellent AND I = High THEN LA = Large
2. IF CS = Fair AND I = Medium THEN LA = Small
3. IF CS = Poor THEN LA = Reject

Crisp Inputs: CS = 650 (μ(Fair) = 0.8, μ(Good) = 0.2), I = 50,000 (μ(Medium) = 0.9) Fired Rules:

• Rule 2: min(0.8, 0.9) = 0.8 → LA = Small (weighted)
• Rule 3: μ(Poor) = 0 → no firing Defuzzification (Centroid):
• Suppose Small = 50,000; Medium = 100,000; Large = 200,000.
• Output = (0.8 × 50,000) / 0.8 = 50,000 (approved loan).

---

8. Advantages and Limitations

Advantages:

• Handles uncertainty: Works with imprecise or noisy data.
• Human-like reasoning: Mimics how humans make decisions (e.g., "kind of hot").
• Real-time adaptability: Used in control systems (e.g., anti-lock brakes).
• Reduces complexity: Avoids precise mathematical models.

Limitations:

• Rule explosion: Too many rules can slow down the system.
• Tuning difficulty: Membership functions and rules require expert knowledge.
• Black-box nature: Hard to explain why a decision was made (unlike crisp systems).

---

Exam Tip

1. Understand the Core Concepts:

   • Memorize the difference between crisp and fuzzy sets.
   • Know the 4 steps of fuzzy inference (fuzzification, rule evaluation, aggregation, defuzzification).
   • Recall defuzzification methods (centroid, max-membership).

2. Practical Questions:

   • Expect worked examples (e.g., given temperature/humidity, calculate fan speed).
   • Be ready to draw membership functions (triangular/trapezoidal) and explain rule firing.
   • Compare fuzzy vs. traditional expert systems in a table.

3. Application-Based Questions:

   • Link fuzzy logic to real-world systems (AC, washing machines, traffic lights).
   • Example question: "A fuzzy expert system is used to control a water heater. The input is water temperature (Cold, Warm, Hot) and the output is heater power (Low, Medium, High). Define 3 fuzzy rules and show how a crisp input of 45°C is processed."

4. Common Pitfalls:

   • Miscounting rule strengths: Always use min for AND, max for OR.
   • Incorrect defuzzification: Double-check centroid calculations.
   • Overcomplicating rules: Stick to 5–10 rules for exam questions.

---