AI technology prevents pump failure

Each year, SA Water invests millions of dollars maintaining critical assets that provide trusted water and wastewater services for its customers. While this has previously been based  on a fixed, time-based preventative maintenance model, Water Engineering Technologies’ (WET) condition monitoring team has a program of work underway to shift to predictive condition assessments of critical assets using smart technology.

Late last year, we installed vibration and temperature sensors at one of the Swan Reach water pump stations, as part of a smart maintenance program proof of concept.

The team wanted to test how well data flowed from smart sensors to SA Water’s statewide, industry-leading monitoring system and machine health software, to help proactively identify and predict any potential failures.

Overseen by our condition monitoring team, we developed a new machine learning solution, applying best practice artificial intelligence modelling and historical data to build intelligence, diagnosing rotating asset fault patterns.

In April 2020, the machine learning solution alerted the team to a potential issue at the pump station. Our project engineer Kaushik Sesha Dollaiah notified our subject matter expert and Lead Condition Monitoring Engineer Stephen Moore to investigate.

With support from Maintenance Supervisor Matt Palmer and his team, Stephen confirmed there was a problem with the pump, and worked to rectify the issue.

"Previously, we would routinely check the pump every three months using a portable analyser. Without the use of smart technology, the issue may not have been picked up for another three months, as the pump had only recently been checked," said Stephen.

Manager Reliability Integration and Condition Monitoring Harvey Pantow added that “by predicting the fault early, we avoided any potential catastrophic failures and further degradation of the asset, as well as impacts to water services for the outer-metropolitan Adelaide area from an unplanned shutdown which typically would incur more costs.”

WET’s performance and condition-based maintenance program enables SA Water to maximise asset lifecycle and reduce operational costs, through the application of data-driven innovative reliability techniques.

In the near future, we will be looking to roll out smart sensors across SA Water’s regional critical pump sets, to provide predictive maintenance analysis and continued monitoring. Moving to this approach provides continued reliable operation of this critical infrastructure, reducing the need for unplanned maintenance and enabling improved safe work practices.

Combined with the optimisation of work management planning and effective work delivery, this will see us work smarter, improve the quality of our maintenance services and free up our people to work on capital projects to achieve ever more reliable water and wastewater services for South Australians.

Diagnostic and prognostic maintenance

Data Analytic Engineer Kaushik Sesha Dollaiah describing data outputs generated by our machine health program.