API 670: The Machinery Protection Standard for Critical Rotating Equipment

API 670 is the standard that defines machinery protection systems for critical rotating equipment such as turbines, centrifugal compressors and large pumps. Where a general vibration survey tells you a machine's health over weeks, an API 670 system watches it continuously and can trip the machine in milliseconds to prevent a catastrophic failure.

What API 670 protects

The standard specifies the sensors, monitors, wiring and logic for a permanent protection system on machines that are too expensive or too dangerous to run to failure. It is the companion to the machinery standards like API 610 for pumps and API 617 for compressors, and it exists because a single wrecked compressor rotor can cost more than the entire monitoring system many times over.

What it measures

• Radial shaft vibration: measured with eddy-current proximity probes that read the shaft's relative displacement in micrometres, usually two probes per bearing at 90 degrees.
• Axial position: proximity probes watch the shaft's thrust position to catch thrust-bearing failure before metal touches metal.
• Bearing temperature: embedded sensors track journal and thrust bearing temperature.
• Speed and overspeed: dedicated sensors detect overspeed, one of the fastest routes to a burst rotor.

How it decides to trip

API 670 defines alarm and shutdown (danger) setpoints for each measurement. To avoid tripping a critical machine on a single faulty sensor, the standard supports voting logic, commonly two-out-of-three, so two independent channels must agree before a shutdown. It also specifies field wiring, power supply redundancy, and rigorous testing so the system fails safe.

Where it fits with routine monitoring

API 670 is protection, not prediction: its job is to stop a machine that is already in danger. It works alongside condition monitoring that trends the same machine against ISO 10816-3 vibration severity to catch slow problems like unbalance and developing bearing failure long before they reach a trip level.

Turning protection data into planning

A protection trip is a last resort; the goal is to never reach it. The continuous shaft-vibration and temperature data an API 670 system produces is exactly the signal a maintenance platform can trend to schedule work before a machine ever approaches its danger setpoint. Fabrico reads those signals from the line and routes the job automatically, turning protection data into planned maintenance rather than an emergency. Book a Fabrico demo to see the closed loop, or read breakdown vs preventive maintenance.

Frequently Asked Questions

What is API 670 used for?

It defines the permanent machinery protection system, sensors, monitors and shutdown logic, for critical rotating equipment such as turbines, compressors and large pumps.

What sensors does API 670 use?

Mainly eddy-current proximity probes for radial shaft vibration and axial position, plus bearing temperature sensors and speed and overspeed detection.

What is voting logic in API 670?

It requires more than one channel to agree before a shutdown, commonly two-out-of-three, so a single failed sensor cannot trip a critical machine unnecessarily.

Is API 670 the same as condition monitoring?

No. API 670 is protection that trips a machine in danger, while condition monitoring trends a machine over time to predict and prevent problems before they reach a trip level. Most critical machines use both.