Smart Networks Shed New Light on Old Water Loss Facts
This is the first post in a new series we’ll call “CTO Insights” by Haggai Scolnicov, TaKaDu’s CTO.
At the recent Water Loss UK workshop and seminar – an interesting, edifying event, as always – I told some of the world’s leading water loss experts what I know about leaks. My talk was clearly “selling ice to Eskimos”, or (closer to the very pleasant Birmingham venue) “bringing coals to Newcastle”.
Really, I just wanted to tell everyone 6 facts about leakage which are not widely known, or just don’t get mentioned enough. At TaKaDu, we’ve been finding leaks and other network faults in customers’ data for several years now, so we have thousands and thousands of individual events to study, each conveniently recorded with the relevant sensor and operations data. This “gallery of leaks” is probably unique. Equally unique, is TaKaDu’s fully-automated statistical analysis of flow and pressure data. To develop and constantly refine this, we have had to study the finest details of a leak’s lifecycle and of the networks we monitor. Only through such study can we help analysts find leaks early, accurately, and reliably, despite the many factors which make this many times harder than theoretical or classroom examples. Look for my list of “How leaks hide in data”, as well as a cheap shot at typical water loss conference slides.
The upshot of all this is that we’ve been able to look at how leaks start, develop, and get repaired, and we noticed (amongst other observations) these 6 interesting and useful facts about “typical leaks”, all detailed and demonstrated in the slides.
- Many leaks start abruptly at 0.5-5 l/s – so individual leaks can have significant impact on water loss, they do not “start at 0”, and there is an abrupt start visible in flow data.
- Leaks start small and grow – at least in many of the leaks we observed, so a 0.5 l/s leak is not “too small to bother”, it is just a 5 l/s leak waiting to happen; repairing a reasonably small leak shows (statistically) better active leakage control than repairing a large one!
- … Or they cause major visible bursts – Many visible bursts develop from a slow leak lasting weeks or months, so with good detection there is time to intervene and prevent the more expensive damage and repairs.
- Leaks rarely last for years – at least the noticeable ones from around 1 l/s and up, so perhaps “background leakage” is just a convenient myth? I asked my audience whether anyone had seen solid evidence for DMAs where “tiny leaks” accounted for any sizeable fraction of the estimated Non-Revenue Water figures, but it seems that background leakage is always assumed, never detected…
- Many leaks last weeks to months – after which they are repaired following a visible burst or through active leakage detection. This accounts for a huge part of the total water loss, and is almost always severely underestimated by water utilities, because of not realising how long the leaks run before an analyst or engineer becomes aware of them. Although a good leakage control program may keep the average flows supplied from increasing over the year, these long “bumps” in the graph can account for most of the utility’s water loss!
- Some (less common) leaks grow very slowly – increasing gradually over many months. Our customers have indicated that this may be a particularly hard to notice and hard to locate type of leak, perhaps physically different to others, and especially valuable when detected by TaKaDu.
I was half-expecting to be told off during the tea break for trying to teach the leakage experts about leaks rather than talking about software, which is what I am supposed to know about, not to mention contradicting or questioning some of the industry’s most cherished assumptions. Instead, I found a line of thoroughly intrigued practitioners and experts, excited by the potential of learning from a “gallery” of real leaks’ hard data. Some of them went so far as to say they had always doubted this or that “standard assumption” about leaks, but never found any decisive evidence until now. A few have since asked for some of this material, to help spread the word.
Here is what I had to say at the workshop, but I heartily recommend browsing the other presentations, available on the event website.