Illustrative Shaft-Order Sampling Exercise

Illustrative Shaft-Order Sampling Exercise

animation
analytics-ml
edge
compression

Scenario: A manufacturing facility wants to examine illustrative shaft-order components on a motor running at 1800 RPM. These examples are possible indicators, not universal bearing-defect frequencies; actual BPFO/BPFI require bearing geometry.

Given:

  • Motor speed: 1800 RPM = 30 Hz (revolutions per second)
  • Illustrative shaft-order components: 1×: 30 Hz (can indicate imbalance); 2×: 60 Hz (can indicate misalignment); 3×: 90 Hz (can indicate looseness); 5×: 150 Hz; 7×: 210 Hz. These are example shaft harmonics, not universal bearing-defect frequencies. Actual BPFO/BPFI require bearing geometry.
  • Highest illustrated component: 210 Hz (7th shaft order).

Question: What is the Nyquist minimum for the highest illustrated component, and what practical rate meets the example’s 2.5× margin?

Solution:

Step 1: Apply Nyquist theorem

Minimum sampling rate = 2 × highest frequency

f_sample_min = 2 × 210 Hz = 420 Hz

Step 2: Calculate practical sampling rate with safety margin

For this example, use 2.5x to 3x the Nyquist minimum as an anti-aliasing filter margin.

f_sample_recommended = 2.5 × 420 Hz = 1,050 Hz
f_sample_practical = 3 × 420 Hz = 1,260 Hz

Round up to a convenient rate that meets the stated 2.5× margin: 1,280 Hz.

Step 3: Check the below-target 1,000 Hz illustration

At 1,000 Hz, the Nyquist frequency is 500 Hz, above the illustrated 30–210 Hz components. This is above the Nyquist minimum but below the example’s 1,050 Hz practical target. Out-of-band input still requires adequate anti-aliasing filtering.

Step 4: Calculate data volume for the separate 1,000 Hz illustration

Single sensor: - Sample rate: 1,000 Hz - Data size: 2 bytes per sample (16-bit ADC) - Data rate: 1,000 × 2 = 2,000 bytes/sec = 2 kB/sec - Daily data: 2 kB/sec × 86,400 sec = 172.8 MB/day

For 100 motors: - Daily data: 100 × 172.8 MB = 17.28 GB/day - Monthly data: 17.28 × 30 = 518.4 GB/month

Step 5: Apply edge FFT compression to the separate 1,000 Hz illustration

Rather than stream raw waveforms to cloud:

# Pseudocode for edge processing every 10 seconds
import numpy

sample_rate_hz = 1000
window_s = 10
samples_per_window = sample_rate_hz * window_s  # 10000

# Perform FFT
fft_result = numpy.fft.rfft(samples)  # 5001 complex bins

# Extract illustrative shaft-order bins
illustrative_freqs_hz = [30, 60, 90, 150, 210]
bin_width = sample_rate_hz / samples_per_window  # 0.1 Hz
selected_bins = [int(f / bin_width) for f in illustrative_freqs_hz]
# Result: bins [300, 600, 900, 1500, 2100]

# Pseudocode: quantize five calibrated magnitudes to uint16, with declared scale,
# saturation handling, and receiver metadata for the selected frequencies.
calibrated_range = ...
magnitudes_u16 = quantize_to_uint16(
    [abs(fft_result[b]) for b in selected_bins], calibrated_range
)
compressed_data = serialize_uint16_with_metadata(
    magnitudes_u16, illustrative_freqs_hz, calibrated_range
)
payload_bytes = len(magnitudes_u16) * 2  # 10 bytes, excluding framing/metadata
compression_ratio = (samples_per_window * 2) / payload_bytes  # 2000x by payload bytes

Step 6: Calculate bandwidth savings

Without compression: - 100 motors × 0.1728 GB/day per motor = 17.28 GB/day

With edge FFT compression (2,000x, assuming five calibrated 16-bit magnitude bins per 10-second window): - 17.28 GB/day / 2,000 = 0.00864 GB/day = 8.64 MB/day

Cost savings:

  • Cloud ingress: $0.09/GB (exercise assumption)
  • Uncompressed: 17.28 GB/day × $0.09/GB = $1.56/day = about $568/year
  • Compressed: 0.00864 GB/day × $0.09/GB = $0.00078/day = about $0.28/year
  • Savings: about $567/year, excluding framing, storage, and processing costs

Key Insight: Sample above the Nyquist minimum with room for anti-aliasing filter roll-off; 2.5–3× the Nyquist rate is a useful starting margin for this example, subject to filter design. Sending selected FFT bins greatly reduces payload while retaining those selected spectral features. Validate fault detection on representative traces because discarded bins and time-domain detail cannot be recovered. The edge gateway computes the features; the cloud receives the chosen diagnostic values.

Back to the chapter