Maira Alvi's deep learning experience

The role of Artificial Intelligence (AI) and deep learning within the water and wastewater industries continues to grow in prominence. One researcher in Western Australia is enhancing treatment efficiency and management of wastewater by integrating AI and deep learning into these processes.

Maira Alvi is a researcher at the University of Western Australia (UWA). Her project, “Deep Learning for Prognostics of Wastewater Treatment Facilities,” won her the Student Water Prize at the Australian Water Association (AWA) Awards.

This project developed a cost-effective prognostic framework using machine learning to forecast key parameters of wastewater processes, enabling better decision support and preventing failures. An automated state extraction enhances system insights for monitoring remote-site events. The framework improves treatment efficiency, advancing knowledge in sustainability and resource optimisation.

“I wanted to enter the water industry to contribute to green technology,” Alvi said. “Another key factor was the industrial PhD opportunity I received through The University of Western Australia and Urban Utilities.”

Originally from Pakistan, Alvi completed her master’s at the University of Engineering and Technology (UET) Lahore. She then moved to Dubai with her husband before relocating to Western Australia to complete her PhD.

“Both of us wanted to complete our doctorates,” she said. “We believed we would benefit from completing our higher education in a multicultural environment among the top one per cent of universities worldwide. UWA has an excellent research environment, so we moved to Perth in 2017.”

Computer science

Through her Ph.D. work, Alvi has been involved in the water and wastewater industry. In her opinion, computer science supports these facilities in many aspects, from data monitoring to predictive maintenance.

“It can be implemented in a range of areas, from data monitoring to analytics, and from predictive maintenance to automation,” said Alvi. “Computer science holds a lot of potential for the industry, particularly with the recent advancement of artificial intelligence.”

Alvi’s research work supports her thinking. It was part of a larger research package at UWA funded by a Cooperative Research Centres Projects (CRC-P) grant. This developed low-cost solution seeks to transform inefficient sewage treatment ponds into self-contained environmental assets benefiting regional communities. It has strong market potential worldwide due to its ability to retrofit existing assets.

The integrated technology can potentially recover water and valuable nutrients suitable for local agricultural uses with minimal odour and greenhouse emissions. A robust control and monitoring system will be incorporated, modernising regional treatment plant operations.

“My PhD research was a sub-project of this CRC-P that sought to transform wastewater treatment in regional Australia,” said Alvi. “It was a collaborative initiative involving key stakeholders such as Urban Utilities (Qld), Power and Water Corporation (NT), Lockyer Valley Regional Council, Aquatec Maxcon, and Queensland Department of Environment and Science. We were specifically interested in how deep learning and artificial intelligence could provide benefits in terms of resource optimisation and improved treatment efficiency and monitoring of wastewater facilities.”

Alvi has long wanted to support green technology and environmental sustainability. With her background in computer science, she sought to provide a different perspective to the project.

“Deep learning has provided answers across many different domains, including autonomous driving,” she said. “I was interested in what I could do and what benefits I could extract for the wastewater industry, which is essential and impacts everyone.”

Findings and challenges

Focusing on environmental sustainability and resource optimisation, Alvi’s research has discovered novel ways to improve the efficiency of wastewater treatment plants. One of her projects uses deep learning models to replace traditional laboratory analysis, providing real-time results and eliminating the need for labour-intensive and time-consuming lab work. This advancement can enhance treatment efficiency and enable real-time adjustments to treatment targets, potentially saving both capital and labour.

Additionally, Alvi’s work features a forecasting model capable of predicting wastewater parameters ahead of time, aiding compliance checks and management decisions. Moreover, an automated state estimation model summarises system dynamics and can anticipate remote site events, enabling timely interventions. These advancements directly enhance the efficiency and sustainability of wastewater treatment processes, aligning with Alvi’s vision of transforming traditional treatment facilities into smart, proactive systems.

“It’s really important to work together and engage in multidisciplinary collaboration,” Alvi said. “It has helped me expand my research endeavours, emphasising the value of diverse perspectives and expertise in pushing the boundaries of knowledge and innovation.”

Alvi identified several data-centric challenges in the wastewater industry, notably the scarcity of quality datasets. She attributed this scarcity to the high costs of sensors and the need for skilled professionals for offline laboratory analyses. Alvi emphasised the significance of this challenge, stating that it hinders deep learning’s potential due to its data-intensive nature. Moreover, she highlighted issues such as imbalanced data due to seasonality, anomalies, and the absence of benchmark or public datasets. These challenges impede the effective application of deep learning in wastewater treatment processes.

Explainability is another aspect Alvi believes is essential and central to model deployment. Also known as eXplainable AI (XAI), it ensures humans can retain intellectual oversight over AI systems, focusing on the reasoning behind the AI’s decisions or predictions, making them more understandable and transparent.

“By increasing our understanding between inputs and outputs, this method opposes the black-box tendency of machine learning,” Alvi said. “We always want to know that given this input, we should expect this output, but we also want to understand why this is occurring.”

Resilience in modelling

Alvi found that predictive models would change from expected outcomes by adding noise.

“These deep predictive models need to be robust to noise and resilient to small changes,” she said. “This is especially important when we don’t have control over the system’s input. When unexpected events occur, we need to assess them to understand what and why it is happening. These models should be able to generalise across different scenarios within the industry. Robustness is crucial as we move forward.”

Alvi’s research has identified opportunities to enhance the resilience and robustness of predictive models in wastewater treatment processes. By addressing the challenges of noise and unexpected events within these models. Her findings suggest that developing models with greater robustness and generalisation capabilities can benefit the industry. However, realising these benefits requires ongoing effort and collaboration across disciplines. Alvi emphasises the importance of not only building resilient models but also ensuring their explainability, enabling broader engagement and cooperation between stakeholders from various fields.

“I want to have a model that does not behave differently with any small tweaks,” Alvi said. “I’m also interested in the explainability of the models. It’s important to be able to take these models to people in other disciplines and have them understand what is happening. That way, they can work with these models and look at them from their perspective. It can help us all work together to improve the deep learning models and algorithms.”

Alvi emphasises the importance of building resilient models and ensuring their explainability, which will enable broader engagement and collaboration among stakeholders from various fields.

Winning and the future

Alvi believes that winning the Student Water Prize of the Year award is a significant recognition of her efforts over the last few years.

“This award holds significant meaning for me as it acknowledges the efforts I’ve dedicated to my work over the past few years,” she said. “It also motivates me to continue working in this field or on similar projects. It means a lot to me, and I was so happy to win the award. I’m just excited that my work received this much recognition.”

Looking ahead, Alvi plans to continue collaborating with industry partners and government agencies throughout her academic career. She aims to provide benefits to the wider community through practical applications of her research.

“If I can positively impact environmental sustainability, I believe that it could be really useful for both the industry and society,” she said. “Collaborating with government and industry is a great opportunity to use my core computer science skills for the benefit of all.”