“Machine failure,” says Marc Hunter, of Invensys global alliance marketing, “is a process, not an event.” He described the process using the unfolding of a car wreck event as an analogy. Traditional vibration monitoring devices, he said, start picking up problems at the grinding and gnashing stage, just before the wreck occurs. Stress wave analysis, on the other hand, can pick up problems far earlier in the failure cycle.
Invensys has partnered with SwanTech, a Curtiss-Wright company, to private label SwanTech’s patented stress wave analysis devices and systems and bring them into the company’s I/A and InFusion framework.
Stress wave analysis uses a passive piezoelectric sonar sensor to monitor ultrasonic energy emitted from rotating machinery. “It is the stethoscope,” Hunter said, “capable of ‘hearing’ friction when it first starts.” Up to eight sensors connect to a SwanGuard device, which does data reduction and sends the data, either wirelessly or over Ethernet or other connection to a SwanServer, a Linux-enabled webserver that does final data analysis and provides a bidirectional link, via OPC, to an I/A or InFusion system.
The system permits the user to see either real-time trending, FFT spectral analysis or histogram analysis to determine likely failure points before they happen. Real-time trend shows instantaneous spikes in the stress wave energy data. With spectral analysis, healthy rotating components produce a flat spectrum and repetitive shock and friction events show up as significant spectral lines. In histogram analysis, normal distribution with a narrow base and few outliers is characteristic of consistent operation. If the distribution is skewed and the base broadens, this is indicative of random friction or shock events of varying intensity—and is almost always a lubrication issue.
John Gager, of Swantech, presented seven case studies, which showed directly the kinds of defects stress-wave analysis can detect, and what the potential cost saving can be from using the methodology. He showed data from a pulp mill in which the mill predicted a cost saving of in excess of a million dollars by installing the system on a single pulper.
Sensors can be mounted in several different configurations, including epoxy-mounting. Gager noted that each sensor can be mounted up to 300 feet away from its SwanGuard receiver.
Hunter described the way the SwanServer connects by OPC to an I/A or InFusion historian. “You can bring down other process data into the SwanServer,” he said, “so you can see what the stress wave energy data means, graphed against things like rotational speed, or throughput. That way the data is even more meaningful. You can bring data up to the historian, or to the alarm or asset management system via OPC as well.”
While the SwanTech devices are slightly more expensive than competitive traditional condition-monitoring systems, on a per-rotating-machine basis, the value proposition isn’t apples-to-apples. The value proposition is that the SwanTech system saves even more money because it can catch problems well in advance of its traditional competition, Hunter said.