By Chris Webb
Condition monitoring (CM) has been at the heart of plant health checks for a couple of decades or more, but the advent of more sophisticated analysis tools looks set to propel the technology by a quantum leap.
By the turn of the millennium, and perhaps before that, genuine artificial intelligence systems will have taken over the traditional role of the experienced technical guru in data interpretation. Whether we like it or not.
Able to learn rapidly and adapt to the foibles and peculiarities of plant operation, these will be capable of analysing huge amounts of data and of predicting with increasing certainty any deterioration or imminent failure in the plant inventory. The potential benefits are clear and elevate the ideology of CM to a higher plane.
An example: trial systems used by ERA Technology use multiple parameter information and make use of a neural network’s ability to learn and adapt solutions to complex data analysis problems.
The nature of the technology allows easier transfer of the techniques to different plant and reduces the amount of effort required to interpret CM information.
We were promised the technology in the 1980s, observes Dr John MacIntyre, who is director of the Centre for Adaptive Systems (CAS) at the University of Sunderland. `Expert systems were hyped in this area (CM), but they didn’t catch on.’ Industry lacked the confidence in what was, says Dr MacIntyre, a technology which failed to measure up to its claims. Consequently, developments in CM returned to what the industry knew best: how to improve signal processing, etc. Now things are about to change.
The CAS is a busy department, with numerous research projects looking at neural nets, fuzzy logic, genetic algorithms, etc, with partners across Europe.
There’s VISION, looking at advanced off-line applications of the technology, with partners from the UK, France, Greece and Finland, and commercial participants such as Nestle and ABB. Then there’s Neural-Maine, the on-line initiative for high capital value plant, mainly involving the UK and Holland. Interested parties include Royal Mail and a number of Dutch power utilities. Again, there’s ATLAS, funded by the UK’s Energy Efficiency Office looking at steam leak detection using systems with artificial intelligence.
The days of reliance on the `expert eye’ to analyse and interpret data, are numbered.
It is, however, one of those disarming anomalies that while CM finds application in a wide range of industrial sectors where its benefits are well known, it is notably absent elsewhere. Here, technology relies heavily on the use of portable vibration monitoring equipment using techniques that have been around for years.
On-line for critical plant
There will always be a market for this, of course, and there will simply be some industries and applications where off-line measurement is more economic and a profoundly logical solution.
On-line CM technology will for the time being be targeting industries with highly critical plant – petrochemicals, oil and gas, etc.
Entek IRD marketing manager Stephen Williams observes a marked shift towards an integrated approach to software systems in CM.
To this end, the company supports implemention of the specifications developed by MIMOSA (Machinery Information Management Open Systems Alliance), an advocate of a more open systems architecture.
Diagnostics traditionally used in CM – vibration and acoustic monitoring, oil analysis, thermography, dye penetration, temperature and pressure monitoring, etc – continue to spawn a wealth of new products.
Measurement of the overall vibration level in terms of an RMS or peak value as the monitored parameter is still one of the simplest methods. Spectral analysis and fast fourier transform (FFT) are more powerful indicators of machinery condition across a range of possible faults – misalignment, bearing wear, imbalance and mechanical looseness.
Similarly, bearings are also monitored using the shock pulse method, while oil analysis is available as an on-line option where the high cost of the equipment can be justified.