feat: Add MLP surrogate with Turbo Mode for 100x faster optimization
Neural Acceleration (MLP Surrogate): - Add run_nn_optimization.py with hybrid FEA/NN workflow - MLP architecture: 4-layer (64->128->128->64) with BatchNorm/Dropout - Three workflow modes: - --all: Sequential export->train->optimize->validate - --hybrid-loop: Iterative Train->NN->Validate->Retrain cycle - --turbo: Aggressive single-best validation (RECOMMENDED) - Turbo mode: 5000 NN trials + 50 FEA validations in ~12 minutes - Separate nn_study.db to avoid overloading dashboard Performance Results (bracket_pareto_3obj study): - NN prediction errors: mass 1-5%, stress 1-4%, stiffness 5-15% - Found minimum mass designs at boundary (angle~30deg, thick~30mm) - 100x speedup vs pure FEA exploration Protocol Operating System: - Add .claude/skills/ with Bootstrap, Cheatsheet, Context Loader - Add docs/protocols/ with operations (OP_01-06) and system (SYS_10-14) - Update SYS_14_NEURAL_ACCELERATION.md with MLP Turbo Mode docs NX Automation: - Add optimization_engine/hooks/ for NX CAD/CAE automation - Add study_wizard.py for guided study creation - Fix FEM mesh update: load idealized part before UpdateFemodel() New Study: - bracket_pareto_3obj: 3-objective Pareto (mass, stress, stiffness) - 167 FEA trials + 5000 NN trials completed - Demonstrates full hybrid workflow 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
Antoine
325
.claude/skills/guided-study-wizard.md
Normal file
325
.claude/skills/guided-study-wizard.md
Normal file
@@ -0,0 +1,325 @@
|
||||
# Guided Study Creation Wizard
|
||||
|
||||
**Version**: 1.0
|
||||
**Purpose**: Interactive conversational wizard for creating new optimization studies from scratch.
|
||||
|
||||
---
|
||||
|
||||
## Overview
|
||||
|
||||
This skill provides a step-by-step guided experience for users who want to create a new optimization study. It asks focused questions to gather requirements, then generates the complete study configuration.
|
||||
|
||||
---
|
||||
|
||||
## Wizard Flow
|
||||
|
||||
### Phase 1: Understanding the Problem (Discovery)
|
||||
|
||||
Start with open-ended questions to understand what the user wants to optimize:
|
||||
|
||||
**Opening Prompt:**
|
||||
```
|
||||
I'll help you set up a new optimization study. Let's start with the basics:
|
||||
|
||||
1. **What are you trying to optimize?**
|
||||
- Describe the physical system (e.g., "a telescope mirror", "a UAV arm", "a bracket")
|
||||
|
||||
2. **What's your goal?**
|
||||
- Minimize weight? Maximize stiffness? Minimize stress? Multiple objectives?
|
||||
|
||||
3. **Do you have an NX model ready?**
|
||||
- If yes, where is it located?
|
||||
- If no, we can discuss what's needed
|
||||
```
|
||||
|
||||
### Phase 2: Model Analysis (If NX model provided)
|
||||
|
||||
If user provides a model path:
|
||||
|
||||
1. **Check the model exists**
|
||||
```python
|
||||
# Verify path
|
||||
model_path = Path(user_provided_path)
|
||||
if model_path.exists():
|
||||
# Proceed with analysis
|
||||
else:
|
||||
# Ask for correct path
|
||||
```
|
||||
|
||||
2. **Extract expressions (design parameters)**
|
||||
- List all NX expressions that could be design variables
|
||||
- Ask user to confirm which ones to optimize
|
||||
|
||||
3. **Identify simulation setup**
|
||||
- What solution types are present? (static, modal, buckling)
|
||||
- What results are available?
|
||||
|
||||
### Phase 3: Define Objectives & Constraints
|
||||
|
||||
Ask focused questions:
|
||||
|
||||
```
|
||||
Based on your model, I can see these results are available:
|
||||
- Displacement (from static solution)
|
||||
- Von Mises stress (from static solution)
|
||||
- Natural frequency (from modal solution)
|
||||
- Mass (from geometry)
|
||||
|
||||
**Questions:**
|
||||
|
||||
1. **Primary Objective** - What do you want to minimize/maximize?
|
||||
Examples: "minimize tip displacement", "minimize mass"
|
||||
|
||||
2. **Secondary Objectives** (optional) - Any other goals?
|
||||
Examples: "also minimize stress", "maximize first frequency"
|
||||
|
||||
3. **Constraints** - What limits must be respected?
|
||||
Examples: "stress < 200 MPa", "frequency > 50 Hz", "mass < 2 kg"
|
||||
```
|
||||
|
||||
### Phase 4: Define Design Space
|
||||
|
||||
For each design variable identified:
|
||||
|
||||
```
|
||||
For parameter `{param_name}` (current value: {current_value}):
|
||||
- **Minimum value**: (default: -20% of current)
|
||||
- **Maximum value**: (default: +20% of current)
|
||||
- **Type**: continuous or discrete?
|
||||
```
|
||||
|
||||
### Phase 5: Optimization Settings
|
||||
|
||||
```
|
||||
**Optimization Configuration:**
|
||||
|
||||
1. **Number of trials**: How thorough should the search be?
|
||||
- Quick exploration: 50-100 trials
|
||||
- Standard: 100-200 trials
|
||||
- Thorough: 200-500 trials
|
||||
- With neural acceleration: 500+ trials
|
||||
|
||||
2. **Protocol Selection** (I'll recommend based on your setup):
|
||||
- Single objective → Protocol 10 (IMSO)
|
||||
- Multi-objective (2-3 goals) → Protocol 11 (NSGA-II)
|
||||
- Large-scale with NN → Protocol 12 (Hybrid)
|
||||
|
||||
3. **Neural Network Acceleration**:
|
||||
- Enable if n_trials > 100 and you want faster iterations
|
||||
```
|
||||
|
||||
### Phase 6: Summary & Confirmation
|
||||
|
||||
Present the complete configuration for user approval:
|
||||
|
||||
```
|
||||
## Study Configuration Summary
|
||||
|
||||
**Study Name**: {study_name}
|
||||
**Location**: studies/{study_name}/
|
||||
|
||||
**Model**: {model_path}
|
||||
|
||||
**Design Variables** ({n_vars} parameters):
|
||||
| Parameter | Min | Max | Type |
|
||||
|-----------|-----|-----|------|
|
||||
| {name1} | {min1} | {max1} | continuous |
|
||||
| ... | ... | ... | ... |
|
||||
|
||||
**Objectives**:
|
||||
- {objective1}: {direction1}
|
||||
- {objective2}: {direction2} (if multi-objective)
|
||||
|
||||
**Constraints**:
|
||||
- {constraint1}
|
||||
- {constraint2}
|
||||
|
||||
**Settings**:
|
||||
- Protocol: {protocol}
|
||||
- Trials: {n_trials}
|
||||
- Sampler: {sampler}
|
||||
- Neural Acceleration: {enabled/disabled}
|
||||
|
||||
---
|
||||
|
||||
Does this look correct?
|
||||
- Type "yes" to generate the study files
|
||||
- Type "change X" to modify a specific setting
|
||||
- Type "start over" to begin again
|
||||
```
|
||||
|
||||
### Phase 7: Generation
|
||||
|
||||
Once confirmed, generate:
|
||||
|
||||
1. Create study directory structure
|
||||
2. Copy model files to working directory
|
||||
3. Generate `optimization_config.json`
|
||||
4. Generate `run_optimization.py`
|
||||
5. Validate everything works
|
||||
|
||||
```
|
||||
✓ Study created successfully!
|
||||
|
||||
**Next Steps:**
|
||||
1. Review the generated files in studies/{study_name}/
|
||||
2. Run a quick validation: `python run_optimization.py --validate`
|
||||
3. Start optimization: `python run_optimization.py --start`
|
||||
|
||||
Or just tell me "start the optimization" and I'll handle it!
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Question Templates
|
||||
|
||||
### For Understanding Goals
|
||||
|
||||
- "What problem are you trying to solve?"
|
||||
- "What makes a 'good' design for your application?"
|
||||
- "Are there any hard limits that must not be exceeded?"
|
||||
- "Is this a weight reduction study, a performance study, or both?"
|
||||
|
||||
### For Design Variables
|
||||
|
||||
- "Which dimensions or parameters should I vary?"
|
||||
- "Are there any parameters that must stay fixed?"
|
||||
- "What are reasonable bounds for {parameter}?"
|
||||
- "Should {parameter} be continuous or discrete (specific values only)?"
|
||||
|
||||
### For Constraints
|
||||
|
||||
- "What's the maximum stress this component can handle?"
|
||||
- "Is there a minimum stiffness requirement?"
|
||||
- "Are there weight limits?"
|
||||
- "What frequency should the structure avoid (resonance concerns)?"
|
||||
|
||||
### For Optimization Settings
|
||||
|
||||
- "How much time can you allocate to this study?"
|
||||
- "Do you need a quick exploration or thorough optimization?"
|
||||
- "Is this a preliminary study or final optimization?"
|
||||
|
||||
---
|
||||
|
||||
## Default Configurations by Use Case
|
||||
|
||||
### Structural Weight Minimization
|
||||
```json
|
||||
{
|
||||
"objectives": [
|
||||
{"name": "mass", "direction": "minimize", "target": null}
|
||||
],
|
||||
"constraints": [
|
||||
{"name": "max_stress", "type": "<=", "value": 200e6, "unit": "Pa"},
|
||||
{"name": "max_displacement", "type": "<=", "value": 0.001, "unit": "m"}
|
||||
],
|
||||
"n_trials": 150,
|
||||
"sampler": "TPE"
|
||||
}
|
||||
```
|
||||
|
||||
### Multi-Objective (Weight vs Performance)
|
||||
```json
|
||||
{
|
||||
"objectives": [
|
||||
{"name": "mass", "direction": "minimize"},
|
||||
{"name": "max_displacement", "direction": "minimize"}
|
||||
],
|
||||
"n_trials": 200,
|
||||
"sampler": "NSGA-II"
|
||||
}
|
||||
```
|
||||
|
||||
### Modal Optimization (Frequency Tuning)
|
||||
```json
|
||||
{
|
||||
"objectives": [
|
||||
{"name": "first_frequency", "direction": "maximize"}
|
||||
],
|
||||
"constraints": [
|
||||
{"name": "mass", "type": "<=", "value": 5.0, "unit": "kg"}
|
||||
],
|
||||
"n_trials": 150,
|
||||
"sampler": "TPE"
|
||||
}
|
||||
```
|
||||
|
||||
### Telescope Mirror (Zernike WFE)
|
||||
```json
|
||||
{
|
||||
"objectives": [
|
||||
{"name": "filtered_rms", "direction": "minimize", "unit": "nm"}
|
||||
],
|
||||
"constraints": [
|
||||
{"name": "mass", "type": "<=", "value": null}
|
||||
],
|
||||
"extractor": "ZernikeExtractor",
|
||||
"n_trials": 200,
|
||||
"sampler": "NSGA-II"
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Error Handling
|
||||
|
||||
### Model Not Found
|
||||
```
|
||||
I couldn't find a model at that path. Let's verify:
|
||||
- Current directory: {cwd}
|
||||
- You specified: {user_path}
|
||||
|
||||
Could you check the path and try again?
|
||||
Tip: Use an absolute path like "C:/Users/.../model.prt"
|
||||
```
|
||||
|
||||
### No Expressions Found
|
||||
```
|
||||
I couldn't find any parametric expressions in this model.
|
||||
|
||||
For optimization, we need parameters defined as NX expressions.
|
||||
Would you like me to explain how to add expressions to your model?
|
||||
```
|
||||
|
||||
### Invalid Constraint
|
||||
```
|
||||
That constraint doesn't match any available results.
|
||||
|
||||
Available results from your model:
|
||||
- {result1}
|
||||
- {result2}
|
||||
|
||||
Which of these would you like to constrain?
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Integration with Dashboard
|
||||
|
||||
When running from the Atomizer dashboard with a connected Claude terminal:
|
||||
|
||||
1. **No study selected** → Offer to create a new study
|
||||
2. **Study selected** → Use that study's context, offer to modify or run
|
||||
|
||||
The dashboard will display the study once created, showing real-time progress.
|
||||
|
||||
---
|
||||
|
||||
## Quick Commands
|
||||
|
||||
For users who know what they want:
|
||||
|
||||
- `create study {name} from {model_path}` - Skip to model analysis
|
||||
- `quick setup` - Use all defaults, just confirm
|
||||
- `copy study {existing} as {new}` - Clone an existing study as starting point
|
||||
|
||||
---
|
||||
|
||||
## Remember
|
||||
|
||||
- **Be conversational** - This is a wizard, not a form
|
||||
- **Offer sensible defaults** - Don't make users specify everything
|
||||
- **Validate as you go** - Catch issues early
|
||||
- **Explain decisions** - Say why you recommend certain settings
|
||||
- **Keep it focused** - One question at a time, don't overwhelm
|
||||
Reference in New Issue
Block a user