Technology

TaKaDu's patented approach uses advanced mathematical and statistical algorithms, specifically developed for handling complex and noisy water network data (including extensive cleaning and streamlining of data), to establish the 'normal' behavior of each network element and the network as a whole - taking into account specific attributes of the single sensor or meter, supply zone or DMA - and compare it in real-time with current network behavior.

Any deviations or discrepancies between observed behavior and the expected behavior are analysed by additional algorithms and using stringent tests to validate the event and rule out any benign explanations. This process contributes to a significant reduction in false alarms.

Once an event is identified, hydraulic and mathematical processes are applied in order to establish the event's exact type, location, and magnitude. Based on system configuration, the event is assigned an owner who is then notified on the event via a multitude of interfaces (web, email, and mobile).

Owing to advanced machine-learning techniques, any change in the event is automatically translated into actionable knowledge delivered to the relevant owner.

TaKaDu's Differentiators

TaKaDu is different from other existing approaches and systems. Some key differences are:

• TaKaDu is field-proven as a live online service, and is used daily by hundreds of water network specialists in many utilities worldwide.
• TaKaDu's algorithms are developed (and constantly improved) and tested with a wide range of network structures, topologies, data-structures, transmission frequencies and 'noise' levels.
• TaKaDu does not require a hydraulic model and does not rely on calibrated models. As such, it can be applied also to networks in which hydraulic models do not exist or are not calibrated.
• TaKaDu provides geo-location information on network events, in network sectors or segments comprising two or more measurement points.
• There is no need to add meters or sensors (unless the utility is interested in improving the detection resolutions, which can be done by adding more sensors or loggers).
• TaKaDu is agnostic to meter and sensor make / type, and can use any data made available by the utility.
• TaKaDu provides a constantly updated 'prediction' for any network element or segment, any time.