WATER & WASTE TREATMENT


OVERCOMING DATA SCARCITY Jaunius Urbonas, senior


data scientist, Aiimi, says the road to intelligent


maintenance in the water industry starts and ends with data


to issues as and when they occur. This approach can cause problems for water companies due to the mission-critical nature of the large-scale assets involved in operations, with failures having the potential to pollute the environment and even risk lives. To improve maintenance capabilities and increase efficiency, water organisations need to harness their data and move from reactive to proactive strategies. Telemetry has been used for many years now


T


to improve maintenance. Pumping stations, for example, are equipped with sensors that deliver information about their activities. But to enable more advanced analytics and harness machine learning capabilities, companies must begin to implement a data-first approach. Trying to advance straight to complex tools


and technologies against a sparse data backdrop will not be successful. A more gradual approach is needed that harnesses data science talent and expertise. The main hurdle to overcome is data scarcity.


The UK water industry grew organically, with assets being constructed and placed as needed. This ad-hoc approach was also applied to maintenance, with assets being patched up and repaired as and when required. This resulted in the gradual growth of maintenance data, captured in varying formats and methods. Much historical data may be locked up in


physical documents that need digitising. As any data scientist will tell you, accurate predictive modelling is not possible without consistent data. It’s therefore essential that organisations first understand the data they have before embarking on predictive modelling projects. Like many businesses, water firms are eager to


harness advanced analytics to improve efficiency, reduce environmental impact and raise their bottom lines. Data scarcity can be overcome by sourcing (installing new sensors) and integrating other datasets (opensource and third-party data providers). Immediate business value can also be realised through doing more with the data that’s available, such as the creation of data visualisations and dashboards


he UK water industry has historically worked on an ad hoc and reactive basis when it comes to maintaining assets, responding


that enable employees across the organisation to easily consume it. Take wastewater pumps as an example.


Water companies track the hours pumping stations run, including how many times they have started and how much water they have pumped within a certain timeframe. To increase knowledge of the pump’s behaviour, and move towards predictive insights, a gradual approach is needed. The first step would be to build data


pipelines and a dashboard to visualise this telemetry data alongside other data, and make it easily accessible. Next steps include creating alerts based on engineering knowledge encoded in algorithms (rule based), anomaly detection (using machine learning models and statistical approaches), and collection or curation of target labels for datasets. All this then could be used to create machine learning models which can predict failures and suggest a root cause. The trick is to start simple, and build up to the machine learning stage of the project. Advanced knowledge about assets and the


environment in which they operate is needed to effectively develop algorithms to correctly identify and predict issues. The best way to accomplish this is by embedding a data scientist in the operation and maintenance (O&M) team for a specific asset. It is a mutually beneficial collaboration. For


example, a data scientist will be able to validate and enrich their understanding of data by building a dashboard while the O&M team will get the ability to see the combined data to make better decisions. For example, heavy rain in certain areas will


lead to increased running time, so no action will be required if we see a pumping station is dealing with this. However, if pumps are running all night when there is no rainfall, action must be taken. Advancing the data strategy from this point


will rely on understanding the deeper context of assets, such as their location, topographical features that contribute to or hinder drainage, etc. As more data sources are combined, and more sensors installed, the overall signal from


the asset will become stronger, allowing more accurate algorithms to be developed. The accuracy will also improve using a


continuous feedback loop as more alerts/predictions generated by algorithms and machine learning models are assessed and root causes recorded. This is a never-ending cycle of continuous


improvement, enabling the O&M team to effectively: • See predictions and anomalies


generated by algorithms and ML models alongside relevant data. • Record an initial assessment if the


prediction is completely wrong or request that the asset is checked. • Document in a structured way a follow-


up (root cause) after the asset that generated the alert has been checked/fixed. Once high enough accuracy is reached,


alerts could be set to automatically generate job orders to send personnel to check the asset together with suggestions of the root cause. Many experts in the water industry have


worked for over 30 years and will take their extensive knowledge with them when they retire. Water companies must seek to encode this knowledge in algorithms which can scan data 24/7 and spot the patterns known to be related to issues with the asset. Extreme weather is also increasing the


pressure on water companies. During normal seasonal patterns a single pumping station might send out a few hundred alerts, but with more extreme weather this could easily grow to thousands, an overwhelming number of alerts that have to be assessed manually. This means rule-based algorithms and data visualisation will be essential. A strong data strategy, driven by


integrated data teams, will see water companies make significant operational improvements. It will realise scalable intelligent maintenance capabilities and begin to future-proof their organisations.


Aiimi www.aiimi.com


SEPTEMBER 2021 | PROCESS & CONTROL 29


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56  |  Page 57  |  Page 58  |  Page 59  |  Page 60  |  Page 61  |  Page 62  |  Page 63  |  Page 64  |  Page 65  |  Page 66  |  Page 67  |  Page 68  |  Page 69  |  Page 70