From Data to Decision: Monitoring and Improving Water Networks using AI/ML

The United Nations Sustainable Development Goal 6, states access to safe drinking water and sanitation for all is a human right. Potable and sewer water networks are critical infrastructure to achieve this globally. In Australia, the annual industrial expenditure on potable and sewer water services is around 4.5 billion AUD. Managing and maintaining a calm water network is important from a consumer, environment and government perspective. Over the past few years, major water utilities across Australia have exponentially added data collection through deployment of a wide range of IoT devices across the potable and sewer water networks. However, currently, this data is rarely used to effectively create insights to enable decision support that transforms productivity, reduces cost and minimises risk.

At Spiral Data, we focus on developing Artificial Intelligence (AI) and Machine Learning (ML) tools towards creating these insights and decision support across the water industry. In this talk, we will discuss two important use cases that were designed, developed and deployed using our python base tech stack through our cloud partner (AWS).

1. Identifying blockages in gravity sewer systems Blockages can happen especially in sewer networks and large blockages can have major environmental impact. However, detecting blockages is difficult since there are no current alarms or alerts. In this use case, we will discuss how we use AI/ML approaches to identify blockages in networks that can then raise an alert. This in turn can in turn enable support crew to address the blockage in a timely manner. These alerts can also enable operations teams to effectively plan around their limited budget and human capacity.

2. Identifying high speed pressure changes in potable water network to monitor pipe failures Potable water networks, especially in large metro areas can experience recurrent large pressure changes (transients). This can be due to users (ex: fire testing) or it can be network induced (ex: valves/pump operation). These pressure transients can cause unexpected failures and as a result monitoring them is of pressing importance. However these data are collected at very high rates and as a result it cannot be analyzed at scale by humans. At SpiralData, we are building automated solutions using AI/ML to identify, analyze and group these high speed pressure transients. This in turn becomes a valuable tool to pursue field investigations to identify source and to also plan better for capex expenses.