Machinery Diagnostics - Complete Examples — Development skill for Claude Code

Machinery Diagnostics - Complete Examples

This guide provides step-by-step examples of complete diagnostic workflows using the Machinery Diagnostics MCP Server.

Table of Contents

• Example 1: Simple FFT Analysis
• Example 2: Bearing Fault Detection
• Example 3: ISO 20816-3 Compliance Check
• Example 4: Complete Bearing Diagnosis
• Example 5: Working with Different Segment Durations
• Example 6: Machine Learning-Based Anomaly Detection
• Example 7: Machine Documentation Reader

---

Example 1: Simple FFT Analysis

Objective

Analyze the frequency content of a vibration signal to identify dominant frequencies.

Scenario

A bearing in healthy condition is operating. You want to analyze the baseline frequency spectrum to establish normal operating characteristics.

Real Dataset Information

• Signal: real_train/baseline_1.csv
• Sampling Rate: 97,656 Hz (from metadata)
• Shaft Speed: 1500 RPM (25 Hz) - from metadata
• Signal Duration: 6.0 seconds
• Condition: Healthy baseline
• Note: Bearing type not specified in metadata

Step-by-Step

**Step 1: List available signals**

In Claude Desktop:

List all available signals in the dataset

**Step 2: Run FFT analysis**

Analyze the FFT spectrum of real_train/baseline_1.csv

Note: The LLM will auto-detect sampling rate from metadata (97656 Hz).
If metadata is missing, you must provide sampling_rate explicitly.

**Step 3: Interpret results**

The tool will analyze the signal and return dominant frequencies:

{
  "peaks": [
    {"frequency": 25.0, "magnitude": 0.68},
    {"frequency": 50.0, "magnitude": 0.42},
    {"frequency": 75.0, "magnitude": 0.25}
  ],
  "signal_duration": 6.0,
  "analyzed_segment_duration": 1.0,
  "frequency_resolution": 1.0,
  "sampling_rate": 97656
}

**Important Notes:**

• The analysis uses 1.0 second random segment by default (not full 6.0s signal)
• Random segment extracted for efficiency (different segment each time unless seed specified)
• Frequency resolution: 1 Hz (1/1.0s) - excellent for most diagnostics
• To analyze full signal, specify segment_duration=None
• To get reproducible results, specify random_seed parameter

**Interpretation:**

• 25 Hz: Shaft rotation frequency (1500 RPM from metadata)
• 50 Hz, 75 Hz: Harmonics of shaft frequency (2×, 3×)
• Clean spectrum: No bearing fault frequencies detected

**Step 4: Generate professional HTML report**

Generate FFT report for real_train/baseline_1.csv

This creates:

• HTML file (fft_spectrum_real_train_baseline_1.html) in reports/ directory
• Interactive Plotly chart with zoom, pan, and hover features
• Embedded met