Making Unplanned Downtime A Myth: Why Water Utilities Should Move From Time-Based To Condition-Based Maintenance

By Terri Lewis and Ankit Dhorajiya

All areas of water utilities are running harder than ever; under tighter budgets, aging assets, workforce churn, and unyielding reliability and compliance demands. In that reality, unplanned downtime isn’t just inconvenient; it threatens water quality, public health, and community trust.

Many utilities still default to time-based maintenance (TBM), servicing equipment by the calendar or hours of operation of the equipment. It’s simple, but it often creates two expensive outcomes: over-maintenance (replacing/service too early) and surprise failures (servicing too late for the actual operating conditions). A smarter path is condition-based maintenance (CBM), servicing when the measured condition of the asset justifies it. Done well, CBM reduces costs and makes unplanned downtime the rare exception.

People → Process → Tools (In That Order)

Technology succeeds only when it empowers the people closest to the process. In practice, that means engaging operators, maintainers, and technicians, i.e., the “boots on the ground”, to shape the workflows, data capture, and responses. Put plainly:

• People without process is action without effectiveness.
• Process without people is just a strategy document.

CBM programs thrive when crews help define inspection routines, choose practical indicators, and co-own the digital workflows that turn raw data into action.

What “AssetIQ” Really Means (And Why Low AssetIQ Isn’t A Deal-Breaker)

AssetIQ is a way to describe the data richness of a given asset — how much usable information it produces about its own health and performance. It defines digitally how “smart” the asset is in its ability to convey performance and problems. Four levers matter: fidelity (signals you have), frequency (how often you get them), breadth (how many indicators you can correlate), and context (metadata such as duty cycles, influent conditions, repair history). Even with low AssetIQ, combining modest signals with context (runtime + starts/hour + operator notes) is enough for many AI models to provide actionable insights.

What AI Really Means For Water Utilities

Artificial intelligence (AI) is often misunderstood as futuristic or abstract, but in practice it is already transforming maintenance and operations in the water sector.

Professor Tom Malone of MIT Sloan offers a simple, accessible definition: “Artificial Intelligence: machines acting in a way that seems intelligent.” Peer-reviewed research adds clarity: AI is the development of computer systems capable of tasks that typically require human intelligence — such as perceiving, learning, reasoning, predicting, and planning.

Together, these definitions remind us: AI isn’t magic — it’s a set of tools that mimic how people make sense of data to act faster, safer, and more effectively. There are four basic types of AI (levels):

1. Reactive AI – Responds to current inputs without memory (e.g., flagging leaks from pressure drops)
2. Limited Memory AI – Learns from past data (e.g., pump wear forecasts, membrane fouling prediction)
3. Theory of Mind AI – Future focused, anticipates human intentions (future, operator decision support)
4. Self-Aware AI – Understands its own limits (conceptual: AI that flags low confidence in a forecast)

Most current CBM applications fall in the first two levels — and that’s enough to cut downtime.

Approaches To AI (How It Works)

There are also different technical approaches to AI, which basically explain how they work:

1. Rules Engines – If–then logic (e.g., runtime alarms)
2. Physics-Based Models – First-principles equations (e.g., hydraulic models)
3. Machine Learning – Learns patterns from data (e.g., pump failure prediction, aeration efficiency optimization)
4. Hybrid AI – Combines rules, physics, and ML (e.g., overflow prediction using rainfall, depth data, and operator rules)

Utilities don’t need exotic AI, just the right mix of rules, physics, and ML that fit their AssetIQ.

Concrete CBM Use Cases For Water, Wastewater, And Stormwater

• Rotating equipment: Pumps, blowers, and fans are the heartbeat of treatment plants and collection systems. By combining vibration signatures with motor current data, utilities can forecast bearing wear and shaft misalignment. This enables maintenance teams to schedule repairs during planned outages instead of reacting to sudden failures, avoiding costly bypass pumping and emergency mobilizations.
• Membranes: Advanced neural network (ANN) models can predict when fouling will occur in membrane bioreactors (MBR), ultrafiltration (UF), or reverse osmosis (RO) membranes by analyzing historical flux and feedwater quality. With this foresight, operators can time chemical clean-in-place (CIP) procedures for maximum effectiveness — reducing chemical use, extending membrane life, and minimizing unplanned downtime caused by sudden permeability loss.
• Collection systems: Rainfall and in-sewer depth/flow data can be fed into predictive models to forecast sanitary sewer overflows (SSOs) or combined sewer overflows (CSOs). With early warnings, utilities can reroute flows, stage bypass pumping, or activate storage facilities before the system exceeds capacity. The outcome is fewer regulatory violations, reduced environmental impacts, and stronger public trust.
• Distribution networks: Pressure anomalies are often early indicators of leaks, bursts, or valve malfunctions. Machine learning algorithms detect subtle changes far earlier than traditional SCADA thresholds, enabling crews to pinpoint and fix issues before they escalate into full breaks or extended service disruptions. The payoff is lower non-revenue water losses, faster response times, and improved customer service.
• Process optimization: Aeration is typically the single largest energy consumer in wastewater treatment. AI-augmented model predictive control (MPC) systems continuously adjust blower output to balance dissolved oxygen levels, energy consumption, and effluent quality. This results in lower operating costs, reduced greenhouse gas emissions, and compliance with increasingly stringent nutrient limits — all without sacrificing reliability.
• Backup power: Monitoring fuel levels, battery voltage, as well as alerts and alarms during testing provides insights on whether backup generators will start when needed. For generators used as prime power, real-time monitoring of voltages, currents, coolant, and fuel levels. in addition to alarms and alerts, ensures power will be on when needed.

Implementation Playbook

A successful shift to condition-based maintenance begins with a clear focus on solving real problems. Operations and maintenance staff are critical partners in identifying priorities and outlining process issues, ensuring that new approaches are both practical and embraced in the field. The process works best when it starts small: focus on one or two high-cost failure modes that consistently cause pain. From there, establish a baseline with simple descriptive dashboards that provide visibility into asset health. Once a clear signal is available, add a predictive model to forecast failures before they occur. The insights generated should then be embedded directly into the CMMS or EAM system, turning predictions into actionable work orders. Finally, the workflows must be co-designed with crews, so the tools fit naturally into daily practice and adoption is seamless. The end goal is not technology for its own sake, but the right technology — used and trusted by staff — to solve problems and eliminate unplanned downtime.

The Business Case (And Culture Case)

The business value of condition-based maintenance is clear. By aligning maintenance activities with actual asset health, utilities can significantly reduce costs, avoiding premature part replacements and the overtime that often accompanies emergency repairs. Implementation of CBM has shown double-digit reductions in costs (~30%) and unplanned downtime reduced by up to 70–75%.

Planned interventions also translate into higher availability, as downtime can be scheduled during low-demand periods rather than forced by surprise failures. CBM further strengthens safety and compliance, since operators spend less time reacting to crises and more time managing stable, predictable systems. Finally, the approach drives workforce empowerment: when frontline staff are equipped with tools that elevate their expertise and make their input essential, job satisfaction and retention rise — reinforcing a culture of collaboration and innovation.

Conclusion

Condition-based maintenance isn’t a luxury, it’s the most practical, people-centered way for water utilities to cut costs, reduce risk, and keep systems running. With even modest AssetIQ (data), and with AI tools tailored to the problem, unplanned downtime can move from inevitability to a myth.

References

• Malone, T. W. (2018). Superminds: The surprising power of people and computers thinking together. Little, Brown Spark.
• Ali, A., & Abdelhadi, A. (2022). Condition-based monitoring and maintenance: state of the art review. Applied Sciences, 12(2), 688.
• Chen, L., Wei, L., Wang, Y., Wang, J., & Li, W. (2022). Monitoring and predictive maintenance of centrifugal pumps based on smart sensors. Sensors, 22(6), 2106.
• Hallaji, S. M., Fang, Y., & Winfrey, B. K. (2022). Predictive maintenance of pumps in civil infrastructure: State-of-the-art, challenges and future directions. Automation in Construction, 134, 104049.
• Kim, H., Jung, K. J., Lee, S., & Jeong, E. H. (2024). Rapid response to pressure variations in water distribution networks through machine learning-enhanced data acquisition. AQUA—Water Infrastructure, Ecosystems and Society, 73(7), 1358-1371.
• Abuwatfa, W. H., AlSawaftah, N., Darwish, N., Pitt, W. G., & Husseini, G. A. (2023). A review on membrane fouling prediction using artificial neural networks (ANNs). Membranes, 13(7), 685.
• Stentoft, P. A., Munk-Nielsen, T., Møller, J. K., Madsen, H., Valverde-Pérez, B., Mikkelsen, P. S., & Vezzaro, L. (2021). Prioritize effluent quality, operational costs or global warming?–Using predictive control of wastewater aeration for flexible management of objectives in WRRFs. Water Research, 196, 116960.
• Sarker, I. H. (2021). Data science and analytics: an overview from data-driven smart computing, decision-making and applications perspective. SN Computer Science, 2(5), 377.
• Collins, C., Dennehy, D., Conboy, K., & Mikalef, P. (2021). Artificial intelligence in information systems research: A systematic literature review and research agenda. International journal of information management, 60, 102383.
• Xu, Y., Liu, X., Cao, X., Huang, C., Liu, E., Qian, S., ... & Zhang, J. (2021). Artificial intelligence: A powerful paradigm for scientific research. The Innovation, 2(4).

Terri Lewis is an electrical engineer and technology strategist with more than 30 years of experience in industrial digital transformation. She is currently a MSCE graduate student at University of Central Florida’s Smart Cities program as well as Principal and Strategic Advisor with CereBulb. She’s focused on innovating at the intersection of industry and infrastructure.

Ankit Dhorajiya has a master’s in engineering combined with over 20 years of expertise in AI and industrial IoT. He has extensive experience developing digital platforms that bring predictive analytics and operational insights to water, energy, and infrastructure clients. He has five patents in advanced data analytics for machinery and leads CereBulb, where work focuses on translating complex technical tools into practical, field-ready solutions.