Edge Processing vs. Raw Data Storage: Requirements for Predictive Maintenance

When considering a predictive maintenance system using IoT, a specification even more important than the sensor's catalog specs is "what format the data is transmitted and stored in."

Most mainstream wireless vibration sensors today employ "edge processing" to minimize communication volume and power consumption. However, this very edge processing is the root cause of the "black box" effect that plagues maintenance engineers in the field.

In this article, we explain the specific contents of edge processing and the decisive difference from the "Raw Data" preservation that the wireless vibration meter "conanair" insists upon.

What Exactly Does "Edge Processing (Sending Results Only)" Refer To?

The phrase "performing edge processing on the sensor side" sounds like advanced technology, but the "results" actually transmitted to the cloud or server generally include the following types of information:

All of these are "processed results." Since the amount of transmitted data is overwhelmingly small, system construction becomes easy. However, is a "result" all that is truly needed at the site of predictive maintenance?

Why Edge Processing Leads to a "Black Box"

The biggest disadvantage of edge processing is that the "Raw Waveform Data," which contains the most information, is discarded at the sensor side.

When an alert sounds to notify an anomaly, engineers in the field will invariably want to confirm: "Why was that judgment made?"
"Was it just picking up a sudden burst of noise?"
"Are the periodic peaks characteristic of bearing damage actually appearing?"

In an edge processing system, since the original waveform does not exist, these questions cannot be answered. It is impossible for the engineer to later re-perform FFT analysis for verification. As a result, they fall into a black box where they have no choice but to trust the system's judgment while remaining skeptical.

The Importance of Raw Waveform Data and the "Data Volume" Dilemma

The "weapon" for field engineers to make maintenance decisions with confidence is unprocessed raw waveform data. However, a major technical dilemma exists here.

Normally, to acquire high-frequency vibrations indicating bearing anomalies as "raw waveforms," the sampling rate must be set very high, resulting in an enormous data volume. This volume strains wireless communication bandwidth (such as Wi-Fi) and rapidly consumes the sensor's battery, making practical operation difficult.

"We want to keep the raw data. But the data volume is too large for wireless operation."

Conanair developed the patented "Under-sampling technology" to resolve this contradiction and continue saving "lightweight raw data" while dramatically reducing communication volume.

30-Day Free Trial & Inquiry

Try Conan Air for free for 30 days. Here is what our users are saying:

• Easy installation for immediate use
• No app required; operate via web browser
• No cloud required for fully automatic measurement
• Affordable, including dedicated software

Try equipment anomaly detection and predictive maintenance for free. Please feel free to contact us.

30-Day Free Trial & Inquiry

Try Conan Air for free for 30 days. Here is what our users are saying:

• Easy installation for immediate use
• No app required; operate via web browser
• No cloud required for fully automatic measurement
• Affordable, including dedicated software

Try equipment anomaly detection and predictive maintenance for free. Please feel free to contact us.

前のページ

Preventing Black Box Vibration Analysis: Raw Data & Under-sampling

次のページ

Patented "Under-sampling" Technology: Mechanism and Clear Technical Limits