In the aftermath of the Heathrow Airport closure following a transformer failure at a power distribution site it seems to me that predicting a transformer failure sufficiently far ahead in time, thus giving time for it to be safely shut down before catastrophic failure, would be much better than trying to minimise the effects of a failure after it has happened.
Using a multivariable Operating Envelope model is a low-cost way of using the additional information contained in variable interactions to predict the onset of a subtle change in equipment behaviour, often with sufficient warning time for an equipment Operator to shut-down the equipment and switch to an alternative.
Think of driving a car at steady speed in top gear along a level highway. When the highway starts to go up a long steep hill the engine note of the car will deepen as it is called upon to work harder and eventually the engine will start to shake the car telling you to change down to a lower gear.
Now imagine the same car and the same highway but this time you are towing a caravan. There wont be much difference when the highway is level but when it starts to go uphill the engine note will deepen sooner and so will the shaking indication that a lower gear is necessary. These are “multi-variable interaction effects”.
For a transformer, which we are inclined to think of as a ‘static’ machine, we know that the insulation between windings will steadily break down until it becomes ineffective allowing short circuits to develop between windings generating heat and often causing the transformer cooling oil to catch fire.
What we don’t know is exactly when this will happen.
We know that it will be expensive in emergency manpower after the failure; that we need to have a spare transformer available at all times; and that there is going to be some downtime of the larger process that the transformer is part of (such as Heathrow Airport) which may be very costly.
The solution that PPCL offers is a real-time geometric model of the Operating Envelope including many measured variables so that it can monitor the interaction effects between them and recognise early signs of relationship changes that are likely indications that breakdown is imminent. The model runs on standard PC’s.
Operating Envelope models are multivariable, build upon the process or equipment knowledge of a senior engineer, require no advanced maths or equations to generate and can be tested by replaying past equipment history data of previous ‘events’ through the envelope model to see how much better that past event could have been managed or even avoided had the model been available back then.
There are many other examples of equipment that is considered ‘static’ until it fails. I will be glad to hear anyones suggestion of other examples of ‘static’ equipment that might be better considered as ‘dynamic’.
There is much more information and detail including several webinar recordings at WWW.PPCL.COM