A particular concern that has been noted is that your system seems to require an initial learning period, where it observes and records normal transient pipeline behaviour. During this period, the system is apparently unable to reliably detect leaks. Can you advise how long this period of exposure typically is? For example, if a pipeline came on line with 2 wells to produce 200 MMSCFD, how long would this "learning" period be? Further, if more wells are brought on line over a period of months, does the system have to revert to "learning" mode each time, and how long are these periods for which they are blind to leaks?

There are four learning periods concerning Atmos Pipe:

1. Configuration tuning (pre-delivery). There is the initial learning period at configuration time, carried out by trained RELi personnel, which takes place before the final commissioning, using real pipeline data. This is a one-off period.
2. Maintenance tuning (on request). The second learning period is performed by the maintenance engineers who supervise the pipeline. If a major change takes place in the pipeline, for instance a flow meter is replaced, or the line is pigged, or a different pump is installed, then any of these changes might modify the behavioural profile of the pipeline. The engineers/operators can "retune" Atmos Pipe by issuing the Tune command. The tuning period can range from a few minutes to a few hours, depending on the size of the pipeline and the customer requirements.
3. Quick tune (automatic). This is used internally by Atmos Pipe to cope with specific problems, for instance a change in product phase, a flaring event or a re-routing operation, when an asynchronous events might temporarily change the behavioural profile of the pipeline and lead to false alarms. As such event are classified as transients, then Atmos must adapt to the new envelope and at the end of it adapt back to the previous conditions after the event has ceased to exist. Quick tuning lasts only a few minutes.
4. Continuous tuning (automatic). Atmos calculates continuously the pipeline resistance and the flow difference between inlets and outlets and updates its internal parameters. This takes place very slowly and Atmos is fully sensitive.

It is important to point out that during normal operation after commissioning, only learning 2, 3 and 4 apply. In addition, all learning times are configurable, according to operational necessity and customer wishes. Longer tuning periods lead to more fitting envelopes and therefore fewer (if any) leak alarms. It is also important to point out that model based leak detection systems have little learning capabilities and following pipeline modifications they require changes to the model which can only be carried out by the supplier: This is both expensive and slow (or even impossible if the supplier is gone out of business). Atmos instead adapts itself automatically or at the operator's wishes.

To answer the questions above, if the customer is concerned about any "blind" periods, a quick tune might be all that is required and for a short line (0-20Km) it might only take 15 minutes.