Understanding the API 670 Standard
API 670 (American Petroleum Institute Standard 670: “Machinery Protection Systems”) is the globally recognised industry standard specifying minimum requirements for the vibration, temperature, and position monitoring systems that provide automatic alarming and shutdown protection for critical rotating machinery in the petroleum, chemical, and power-generation industries. It defines sensor types and quantities, alarm and trip setpoints, redundancy requirements, testing procedures, and system-design criteria — all aimed at reliable machinery protection against catastrophic failures. Where general guidance such as ISO 20816 tells you how to evaluate vibration severity, API 670 tells you how to build the permanently installed system that watches the machine and trips it before it destroys itself.
API 670 compliance is mandatory for most large turbomachinery (above roughly 10,000 HP) in hydrocarbon service and is widely adopted as best practice well beyond the petroleum industry. It represents the consensus approach to protecting critical machinery, weighing safety and reliability against practical implementation.
1. Scope and Applicability
The standard targets the machines whose unplanned failure would be most dangerous or most expensive — typically large, high-speed, single-train equipment with no installed spare.
Covered Equipment
- Steam and gas turbines
- Centrifugal and axial compressors
- Centrifugal pumps in critical service
- Generators and motors above ~10,000 HP
- Expanders and blowers
- Generally, critical turbomachinery across petroleum and power
When It Is Required
- Equipment above ~10,000 HP — typically mandatory
- Critical service (no backup, high consequence of failure)
- Contractual requirements between purchaser and vendor
- Corporate engineering standards
- Often applied voluntarily as recognised best practice
2. Key Requirements
API 670 prescribes what to measure, how many sensors to use, and the indicative levels at which the system should act. The headline measurements are radial vibration, axial position, a phase reference, and bearing temperature.
Radial Vibration Monitoring
- Sensors: XY proximity probe pairs at each bearing (a minimum of four probes per machine, two per bearing).
- Measurement: shaft displacement relative to the bearing — the non-contact eddy-current probes watch the shaft itself, not the housing.
- Alarm: typically 10–15 mils (250–380 µm) peak-to-peak.
- Trip: typically 25 mils (635 µm) peak-to-peak.
- Response time: under 1 second from trip detection to shutdown initiation.
Axial Position Monitoring
- Sensors: two axial displacement probes (redundant).
- Purpose: monitor thrust-bearing condition and axial rotor position.
- Alarm/Trip: set according to the available axial clearance.
Phase Reference (Keyphasor)
- Sensors: two keyphasor probes (redundant).
- Purpose: once-per-revolution timing for phase and speed measurement.
- Requirement: mandatory for complete rotor analysis — without it, plots such as Bode, polar and orbit cannot be produced.
Bearing Temperature
- Sensors: two RTDs per bearing (redundant).
- Alarm: typically 95–105 °C.
- Trip: typically 110–120 °C.
3. Redundancy and Voting
A protection system is only useful if it trips for real faults and stays quiet for false ones. API 670 achieves this balance through redundant sensors feeding voting logic.
Sensor Redundancy
- A minimum of two sensors for each critical parameter.
- Prevents a single-point sensor failure from disabling protection.
- Enables voting logic to discriminate genuine events from sensor faults.
Voting Logic
- 2-out-of-2 (AND): both sensors must agree before a trip is issued.
- 2-out-of-3: any two of three sensors trigger action — the preferred arrangement for the most critical machines.
- Purpose: balance nuisance-trip prevention against the need for dependable failure protection.
Monitor Redundancy
- Dual monitor racks are sometimes specified.
- Independent power supplies for each channel.
- Fail-safe design throughout the chain.
4. System Features and Testing
Required Functions
- Real-time displays of all monitored parameters.
- Alarm and trip functions with configurable time delays.
- Alarm acknowledgement and reset.
- Bode and orbit plots for diagnostics.
- Event recording and historical archiving.
- Diagnostic software tools.
Data Recording
- Continuous trending of all parameters.
- Capture of startup and shutdown transients, where many problems first reveal themselves.
- Alarm-event data snapshots.
- Long-term historical archiving.
Acceptance and Periodic Testing
API 670 specifies a structured test programme so that the protection is proven before commissioning and stays proven through life:
- Factory Acceptance Test (FAT): the complete system is tested before shipment — all functions verified, calibration confirmed, documentation provided.
- Site Acceptance Test (SAT): after installation, a full functional test verifies every sensor channel, exercises the alarm and trip functions, and validates the system against its specification.
- Periodic testing: quarterly or annual functional tests confirm the trip circuits still operate, recheck sensor calibration, and keep the documentation current.
5. Revisions, Related Standards, and Field Balancing
The 5th Edition (2014) modernised the standard for digital systems, added cyber-security requirements, updated sensor specifications, and improved the testing procedures; it is the version most widely implemented today. API 670 also sits within a family of related documents:
- API 617: axial and centrifugal compressors.
- API 610: centrifugal pumps.
- API 684: rotor-dynamics analysis.
- ISO 7919: shaft-vibration limits (the shaft-relative counterpart to housing measurements).
- ISO 20816: bearing-housing vibration limits (formerly ISO 10816).
It is worth being clear about what API 670 does not do: it protects a machine, but it does not correct one. When the permanent system flags a rising 1× peak from unbalance, the fix is still a balancing job, usually performed on site. A portable two-channel analyser such as the Balanset-1A complements an API 670 installation in exactly this gap — measuring 1× amplitude and phase in the machine’s own bearings, computing the correction weights, and verifying the residual unbalance after the repair, without disturbing the fixed protection probes.
In sum, API 670 is the cornerstone standard for machinery-protection systems in the petroleum, chemical, and power industries. By specifying sensor configurations, redundancy, alarm and trip levels, and testing, it ensures consistent, reliable protection across facilities worldwide, preventing catastrophic turbomachinery failures through proven monitoring and automatic shutdown.