Mechanical Properties Prediction

# Mechanical Properties Prediction

## Introduction & Motivation

Predicting mechanical properties from material composition and microstructure accelerates materials design for engineering applications. ML models learn structure-property relationships for rapid property screening and materials optimization.

Motivation: Predict mechanical properties for materials engineering.

Applications: Strength prediction, toughness estimation, failure prediction, design optimization.

---

## Core Concepts & Theory

### Yield Strength

Plastic deformation onset.

### Ultimate Tensile Strength

Maximum load capability.

### Hardness

Resistance to indentation.

### Fracture Toughness

Crack resistance.

---

## Mathematical Formulation

Hall-Petch Relationship:
$$\sigma_y = \sigma_0 + k d^{-1/2}$$

Strength Components:
$$\sigma_{ ext{total}} = \sigma_{ ext{solid solution}} + \sigma_{ ext{dislocation}} + \sigma_{ ext{precipitate}}$$

Hardness:
$$H = \frac{2P \sin( heta/2)}{\pi a^2}$$

---

## Advanced Theory & Extensions

### Microstructural Strengthening

Grain size and precipitates.

### Strain Hardening

Work hardening effects.

### Temperature Effects

Thermal softening.

---

## Computational Considerations

Microstructure: O(N_features·D) complexity.

Property Model: O(D²) network.

Prediction: O(D) per composition.

---

## Practical Implementation Strategies

### Composition Encoding

Element fractions.

### Microstructure Features

Grain size, phase distribution.

### Processing History

Heat treatment effects.

---

## Benchmark Datasets & Evaluation

MatWeb: Material properties.

NIST Database: Mechanical data.

Literature Values: Published results.

---

## Key Challenges & Limitations

### Anisotropy

Directional dependence.

### Processing Variation

Manufacturing effects.

### Scale Dependence

Sample size effects.

---

## Hyperparameter Tuning

Hidden units: 64-256 neurons.

Dropout: 0.2-0.4 regularization.

Learning rate: 1e-4 to 1e-2 schedule.

---

## Real-World Applications & Case Studies

Steel: Strength optimization.

Aluminum: Aerospace structures.

Titanium: High-performance applications.

---

## Integration with Other Methods

Mechanical ML + microstructure simulation; + experiments; + design.

---

## Summary & Key Takeaways

ML predicts mechanical properties efficiently.

Principles:
1. Composition: Element encoding.
2. Microstructure: Phase and grain size.
3. Processing: Treatment history.
4. Properties: Strength and toughness.
5. Optimization: Material design.

---

## Appendix: Practical Labs

### Lab 1: Microstructure Feature Extraction

import numpy as np

def extract_mechanical_features(composition, grain_size, phase_fraction):
 """Extract mechanical property descriptors"""
 features = np.array([
 np.sum(composition),
 grain_size,
 phase_fraction,
 1.0 / (grain_size + 1e-6)
 ])
 assert len(features) == 4, "Feature dimension error"
 return features

comp = np.array([0.3, 0.5, 0.2])
features = extract_mechanical_features(comp, 10, 0.6)

assert features.shape == (4,), "Feature extraction failed"
print(f"✓ Mechanical features: {features}")

### Lab 2: Strength Prediction

import numpy as np

class MechanicalPropertyPredictor:
 def __init__(self, descriptor_dim=10):
 self.weights = np.random.randn(descriptor_dim, 3) * 0.1
 self.bias = np.array([200, 400, 30])
 
 def predict_properties(self, features):
 """Predict yield, UTS, hardness"""
 properties = features @ self.weights + self.bias
 return properties

features = np.random.randn(10)
predictor = MechanicalPropertyPredictor()
props = predictor.predict_properties(features)

assert props.shape == (3,), "Prediction failed"
print(f"✓ Yield: {props[0]:.0f}, UTS: {props[1]:.0f}, HV: {props[2]:.1f}")

### Lab 3: Hall-Petch Analysis

import numpy as np

def calculate_yield_strength(sigma_0, k, grain_size):
 """Calculate yield strength from grain size"""
 sigma_y = sigma_0 + k / np.sqrt(grain_size + 1e-6)
 assert sigma_y > 0, "Strength must be positive"
 return sigma_y

sigma0 = 100 # MPa
k = 200 # Hall-Petch coefficient
d = 10 # microns

strength = calculate_yield_strength(sigma0, k, d)

assert strength > 0, "Strength calculation failed"
print(f"✓ Yield strength: {strength:.0f} MPa")

### Lab 4: Property Optimization

import numpy as np

class PropertyOptimizer:
 def __init__(self, target_strength=400):
 self.target = target_strength
 
 def optimize_composition(self, n_iterations=20):
 """Optimize composition for target strength"""
 best_comp = np.random.dirichlet(np.ones(3))
 best_error = float('inf')
 
 for _ in range(n_iterations):
 comp = best_comp + np.random.randn(3) * 0.05
 comp = np.clip(comp, 0, 1)
 comp = comp / comp.sum()
 
 strength = 200 + np.sum(comp * 300)
 error = abs(strength - self.target)
 
 if error < best_error:
 best_error = error
 best_comp = comp
 
 return best_comp

opt = PropertyOptimizer(target_strength=350)
optimal = opt.optimize_composition()

assert np.isclose(np.sum(optimal), 1.0), "Optimization failed"
print(f"✓ Optimized composition: {optimal}")

---