How to detect equipment failures through spectral nalysis

Predictive maintenance is essential for ensuring the reliability of rotating assets, and vibration analysis stands out as one of the most effective techniques for early failure detection.

This method is illustrated in the well-known DIPF curve (Figure 1), which shows the different phases an asset goes through — from design to the end of its useful life.

The DIPF curve expands on the traditional PF curve, adding the design and commissioning phases of equipment.

Why vibration data collection matters

There are several technologies available for collecting vibration data on industrial assets. In this article, we focus on wireless sensors permanently installed on critical components, which provide two main types of data:

• Global (continuous) data
• Spectral data

Global data enables basic monitoring of a machine’s operating levels, quickly highlighting deviations from expected patterns.

For example, using Dynamox wireless sensors, a client observed a clear increase in RMS velocity levels on the axial shaft of an engine, indicating abnormal vibration behavior.

However, while global data is useful for quick detection, it does not reveal the failure mode or root cause. For that, spectral analysis is essential.

How do failures appear in Spectral Analysis?

Spectral analysis investigates signals in the frequency domain using the Fourier Transform, applied to the original time-domain signal (waveform).

The waveform contains valuable information about component condition and is represented by a sum of sine and cosine waves of varying frequencies and amplitudes. After processing, this waveform becomes a frequency spectrum.

Dominant frequencies or energy concentration in specific frequency bands often indicate failures. Each failure type alters the machine’s spectral signature, making it possible to identify the root cause.

Example: Detecting unbalance in rotating equipment

Consider an engine operating near 3600 RPM that showed increased vibration levels in August 2021.

By analyzing the waveform and spectrum from a sensor installed on the coupled side, the root cause was identified: unbalance.

The frequency-domain analysis revealed a strong vibration peak at the rotational frequency (3492 RPM or 58.20 Hz), which is characteristic of unbalance.

This fault typically appears more prominently in horizontal and vertical axes, with minimal activity in the axial axis — exactly what the data showed: where there is a dominance of horizontal vibration (blue), followed by vertical vibration (yellow), and virtually no vibration in the axial shaft (pink).

Additionally, historical analysis demonstrated the progression of unbalance over time through a spectral cascade of RMS velocity.

After exploring unbalance detection, let’s look at other common failure modes and how spectral analysis can reveal them.

Identifying misalignment through frequency peaks

The figure below shows the spectrum at the bearing housing monitoring point of an electric motor. Notice the prominent peak at approximately 40 Hz, which corresponds to 2× the shaft’s rotational frequency.

Although the 1× RPM frequency is also present, its intensity is lower than the 2× RPM frequency. This pattern in the spectrum strongly indicates parallel misalignment on the drive shaft.

Additionally, stresses from the motor’s magnetic field appear as excitations at the rotor groove frequency (1080 Hz) and stator groove frequency (1500 Hz), along with sidebands at the shaft’s rotational frequency.

Gearbox failures: Cracked teeth and harmonics

Another common failure is cracked gear teeth in gearboxes. The severity of this fault can be assessed by the presence of harmonics of the gear frequency, as shown in the figure below.

Cracked teeth also cause modulations in the gear frequency relative to the rotation of the defective gear. This results in sidebands with multiple harmonics around the damaged gear’s rotation.

The Harmonics Marker helps pinpoint these harmonics, with the fundamental frequency at 343.75 Hz.

Similarly, the Sideband Marker reveals that sidebands are spaced at 5.08 Hz, corresponding to the output shaft’s rotational speed — confirming an output gear failure.

What if the fault doesn’t appear in the spectrum?

Sometimes, faults are not clearly visible in the waveform or spectrum, requiring advanced signal processing techniques.

For example, bearing failure frequencies may be masked by other components, making diagnosis challenging.

In the figure below, the spectrum shows multiple peaks and harmonics, but applying the bearing failure frequency marker reveals very low amplitudes at these frequencies.

Using the envelope technique for hidden faults

In such cases, signal demodulation using the Envelope Technique is essential. This method extracts the waveform envelope after applying a bandpass filter. The Dynamox Platform offers different filter frequencies to ensure accurate failure diagnosis.

As shown in the figure, the envelope spectrum reveals harmonics at the outer race bearing failure frequency, indicating the need for continuous monitoring. For more details on bearing fault detection, check out our dedicated article.

The Envelope Technique is available on the Dynamox Platform, along with a database of failure frequencies (BPFO, BPFI, BSF, FTF) for nearly 70,000 bearings of various makes and models. Frequency markers can be added to the spectrum for easier interpretation.

How to set up spectral data collection for accurate diagnosis

Successful spectral analysis depends on proper setup, which impacts signal quality and diagnostic accuracy.

• Dynamic Range: Choose based on expected amplitude levels.
  • Low values can cause signal saturation, hiding fault evolution.
  • High values on low-vibration assets reduce resolution, impairing spectrum quality.
• Sampling Frequency: Determines the maximum identifiable frequency (Nyquist frequency).
  • High-frequency failure modes require higher sampling rates.
  • Lower frequencies allow longer collection times and better spectral resolution.

The combination of sampling frequency and collection duration defines the number of lines in the spectrum.

DynaLoggers: Wireless sensors for predictive maintenance

Dynamox’s DynaLogger family of wireless vibration and temperature sensors is designed for diverse machine types. Technical details are available in the product datasheets.

Dynamox also provides consulting services to help you choose the right sensor for spectral analysis and offers full support for using the Dynamox Platform effectively.

Contact us today for a quote. 

With DynaLoggers, you can collect global levels for continuous monitoring and spectral data for detailed diagnostics. Data is sent to the Dynamox Platform via the App or automatically through a Gateway.

Spectral collection settings include axis selection, sampling frequency, dynamic range, and collection duration.

Once data is available on the Dynamox Platform, users can view spectral analyses alongside continuous monitoring of speed, acceleration, and temperature.

This enables quick diagnosis when trends indicate potential issues, allowing maintenance teams to plan corrective actions efficiently.

Spectral analysis can be applied to motors, gearboxes, bearings, pumps, and more — making it a cornerstone of predictive maintenance strategies.

Key takeaways for maintenance teams

• Use wireless vibration sensors for continuous monitoring of critical components.
• Combine global data for quick detection with spectral analysis for accurate diagnosis.
• Plan corrective actions based on the severity and type of failure identified in the spectrum.

Do you want to learn more about how spectral analysis can be applied in your industrial plant? Talk to a Dynamox specialist.