AeronauticsARTIFICIAL INTELLIGENCE
When machines start talking
ACL Digital, an ALTEN company, built an integrated IoT and AI/ML-based predictive maintenance platform for a major utility provider in the United States, combining sensor analytics and computer vision to monitor electricity, gas and steam infrastructure, above ground and underground, in real time, and act on equipment issues before failures occur.
Utility providers are under constant pressure to keep infrastructure running reliably while controlling operating costs. For companies managing electricity, gas and steam generation at scale, unplanned equipment failures, whether in visible machinery or underground electrical structures, can disrupt service and drive up maintenance costs, making early detection of abnormal behaviour essential to reliable energy delivery. A major utility provider in the United States engaged ACL Digital to help it respond to these needs.
Challenge: Develop a smart system to analyse data from machinery and underground electrical assets, identify abnormal operations and enable preventive maintenance to avoid costly equipment failures.
Solutions: An integrated IoT and AI/ML-based predictive maintenance platform, combining sensor analytics and computer vision, that gathers, processes and analyses both equipment sensor data and thermal imagery to trigger preventive maintenance before failures occur.
Benefits:
- Advanced monitoring and predictive maintenance
- Cost efficiency and resource optimisation
- Advanced diagnostics and resolution capabilities
- Saved up to 50% on maintenance cost
- Faster and more efficient services
- Quick resolutions and high customer satisfaction
- 250K+ underground electrical structures brought into the inspection scope
- 95%+ hotspot detection accuracy from thermal imagery
Key highlights:
- TensorFlow-based ML model trained and validated on compute infrastructure and ported to an embedded NXP hardware platform.
- Edge AI deployment enabling inference closer to the asset, reducing dependency on centralised processing.
- Automated visual inspection of electrical assets, helping reduce manual inspection effort and dependency on specialised field expertise.
- Multiple asset/defect classes potentially identified from thermal imagery.
Two data streams, one platform
The platform draws on two complementary data streams to give the client a fuller picture of asset health. Above ground, ACL Digital built an IoT system on AWS to gather sensor data, including vibration and temperature readings, from machinery via CENTAURI 200 gateways using wireless protocols; AI/ML models then analyse this sensor data to identify abnormal equipment behaviour and trigger preventive maintenance before failures occur. Underground, the platform applies computer vision to thermal imagery of the client’s electrical structures: a TensorFlow-based ML model was trained and validated on compute infrastructure, then ported to an embedded NXP hardware platform, enabling edge AI inference closer to the asset and reducing dependency on centralised processing.
This automates the visual inspection of electrical assets, helping reduce manual inspection effort and the client’s dependency on specialised field expertise, while identifying multiple asset and defect classes from thermal imagery, including early signs of electrical hotspots. To make both data streams actionable, ACL Digital designed a customisable dashboard that lets the client’s teams easily view sensor data and analyse system health with configurable parameters, helping them spot potential issues early across both parts of the infrastructure. The solution relies on the IoTiFy framework, Python, TensorFlow Lite and Scikit-learn, AWS IoT Cloud, Edge ML, and a TensorFlow model ported to embedded NXP edge hardware.
Key performance indicators:
- 250K+ underground electrical structures brought into the inspection scope
- 25K+ thermal images annotated for ML training
- 95%+ hotspot detection accuracy
Reliable energy, lower costs
By catching abnormal equipment behaviour early, whether from vibration signatures above ground or thermal hotspots below it, the platform allows the client to shift from reactive to preventive maintenance, saving up to 50% on associated costs while delivering faster, more efficient services. Quick resolutions and advanced diagnostics capabilities give operations teams the confidence to act before issues escalate, supporting reliable energy delivery and driving high customer satisfaction.
The power of AI has enabled this utility provider to modernise infrastructure monitoring, above ground and below, and cut maintenance costs while keeping energy delivery reliable.
