From 2‑hour gaps to continuous visibility – a case study with real waveform data

1. Demonstrate, using a fault case, the necessity of rapid deterioration fault capture capability.

2. High-density index sampling and waveform additional collection for precise analysis.

Technical Background

• It tends to miss critical deterioration points, and the effective waveform data collected is insufficient to support accurate diagnostics.

To overcome this, we adopt a high‑density index data sampling combined with anomaly‑triggered waveform collection strategy. For sudden deterioration faults:

• High‑density index data sampling obtains multiple sets of high‑density status data in a short time.

• Cross‑validation of multi‑point features significantly reduces the risk of false alarms.

• Even if a single data point shows an anomaly, an alarm can be triggered in real time without waiting for a full sampling cycle.

• After an anomaly is detected, additional waveform collection is automatically triggered to supplement key diagnostic information. This enables a more complete capture of the fault evolution process and provides a solid basis for diagnostic analysis.

This achieves a closed‑loop mechanism of "fast response + accurate confirmation + data supplementation".

The mechanism is particularly suitable for sudden fault scenarios, filling the blind spots of low-frequency monitoring. It ensures alarm reliability while supporting subsequent deterioration path analysis and customer decision‑making.

Case Study (Hydrogen/Oil Combined Workshop – Light Ends Recovery Area / P304B)

The following sudden fault of a centrifugal pump is described in detail as an example:

2.1 Basic Information

Double‑support centrifugal pump with two vibration analysis sensors installed: 3H at the pump drive end and 4H at the non‑drive end. The sensor type is RONDS triaxial wireless sensor.

2.2 Data Analysis

2.2.1 Case Background

• The high‑frequency acceleration RMS value showed a sudden increase within 5 minutes.

• At the same time, the RMS value for both low‑frequency acceleration and velocity also increased. Among these:

     ·The low‑frequency acceleration changed more significantly.

     ·Its peak value reached approximately three times that of the stable period.

• Subsequently, the vibration briefly decreased and then re‑stabilized at a relatively high level.

If relying solely on low-frequency waveform collection, the critical deterioration process of "sudden increase – decrease – re‑stabilization" would very likely be missed.

• A large amount of anomaly‑triggered additional waveform data was captured between 10:00 and 12:00.

• This not only recorded the "result", but also fully reconstructed the entire fault evolution process.

• This provides for subsequent diagnostics:

            ·More accurate fault type identification

            ·More reliable damage severity assessment

            ·More complete basis for evolution path analysis

2.2.2 Detailed Analysis Process

2.2.2.1 High‑Density Index Sampling Enables Capture of Rapid Sudden Faults

A sudden rise in short-term index value was observed: the high‑frequency acceleration RMS values at both the pump drive end and non‑drive end increased significantly within 5 minutes (short‑term index sampling rate: one set every 5 minutes).

The low‑frequency acceleration RMS values and the velocity RMS values at both measuring points on the pump also showed a clear jump, with the low‑frequency acceleration exhibiting a more pronounced change.

From the trend perspective, the low‑frequency acceleration reached its peak within a short time, with an amplitude approximately three times that of the previous stable period. Subsequently, the vibration level dropped significantly and re‑stabilized.

• The low‑frequency acceleration overall amplitude remained slightly higher than that of the initial stable phase.

• The velocity vibration amplitude, however, fell below the previous stable level.

• Inability to fully reconstruct the entire fault evolution process.

• Difficulty in capturing key deterioration points.

• Compromised accuracy of subsequent fault type identification and detailed analysis.

2.2.2.2 Capturing Every Critical Moment of the Fault

2.2.2.3 Seeing the Complete Trajectory of Fault Development

This section compares and analyzes data characteristics such as waveforms, spectra, and envelope spectra at different stages of the fault. Three key time points are selected:

• Early Fault

• Fault Deterioration

• Late Fault

(1) Waveform and spectrum comparison for high‑frequency acceleration:

When the fault begins to deteriorate, the main feature is an increase in the high‑frequency noise floor.

In the late fault stage, the dominant features are rotational frequency and its rich harmonics, while the high‑frequency noise floor disappears.

(2) Characteristic changes in envelope spectra of high‑frequency acceleration at different times:

Early fault: During the stable vibration stage of high‑frequency acceleration, cage frequency and ball spin frequency can be observed in the spectrum following envelope demodulation.

Fault deterioration: At the onset of the fault, the spectrum following envelope demodulation is dominated by noise energy.

Late fault: In the late stage, the spectrum following envelope demodulation is dominated by rich harmonics of the rotational frequency.

(3) Characteristic changes in waveforms and spectra of low‑frequency acceleration at different times

In the late fault stage, the low‑frequency acceleration waveform shows clear periodic rotational frequency impacts, and its spectrum is dominated by the rotational frequency and its rich harmonics.

(4) Temperature trend analysis

As vibration increased, temperature also rose significantly, from approximately 33 °C to 116 °C.

Based on the above waveform data acquired through additional collection, the following observations are made:

• acceleration and velocity RMS values increased sharply;

• the time‑domain waveform showed rotational frequency impacts;

• the spectrum contained numerous rotational frequency harmonics;

• temperature rose continuously and significantly (from about 33 °C to 116 °C).

These indicate severe looseness characteristics.

2.2.2.4 Conclusion and Recommendations

Diagnostic conclusion: Sudden fracture of the pump drive‑end bearing cage. After the cage fractured, the rotor became unstable, causing rubbing, friction, and a rise in temperature.

Maintenance recommendations: Further operation is not recommended. Immediate inspection and confirmation of equipment condition are advised, and the pump drive‑end bearing damage should be checked as soon as possible.

2.3 On‑site Feedback

Based on our diagnostic conclusions and recommendations, a shutdown was arranged for maintenance. After the repair, on‑site feedback confirmed severe damage to the pump drive‑end bearing and a fractured cage.

Closing Remarks

Through the high‑density index data sampling and anomaly‑triggered waveform collection mechanism, the system is able to obtain complete key data during rapid fault evolution.

• High‑density index sampling improves sensitivity to abnormal changes occurring over short time periods.

• Anomaly‑triggered waveform collection captures high‑value analysis data at critical time points, effectively avoiding the missed detections or information loss that can occur with low-frequency periodic collection.

The combination of these two approaches not only improves the timeliness of response to sudden or rapid deterioration faults, but also ensures the continuity and completeness of the fault evolution process, thereby significantly enhancing the accuracy and reliability of diagnostic analysis.