Undetected anomalies grow into big issues

Reactive maintenance practices fail to address technical issues at their inception, often leading to machine failures and unplanned downtime.

Here are just some example of the costs that can arise from undetected anomalies:

  • When Virgin Blue suffered an 11-day unplanned downtime in 2010, it resulted in 50,000 disrupted passengers, 400 grounded flights and $20 million in losses.
  • An average mid-size Liquified Natural Gas facility loses about $ 150 million every year due to unplanned downtime.

  • In the power generation sector, unexpected disruptions cost every year between 3% and 8% of total capacity, resulting in $10 billion lost production revenue.

Catch problems before there is a problem

Even the most robust machine will eventually experience problems. This is normal and should not be feared.

What’s important is how we handle these problems. If we are able to detect them early, we can prevent them from growing into unnecessary failures and downtime.

Early detection allows us to move from a reactive maintenance approach to a proactive one.

This is the essence of ⍺I Predictive Maintenance and early anomaly detection is it’s most important component.

To give you an idea of what can be achieved with proactive maintenance operations, consider the manufacturing industry where predictive anomaly detection results in savings of up to 12% over scheduled repairs, maintenance cost reductions of up to 30% and elimination of breakdowns by up to 70%.

CASE STUDY

Detecting Anomalies in Aircraft Sensor Data

The problem

For an aircraft manufacturer, catching any malfunction of the vehicle’s components quickly is vital. This not only impact safety, but also aircraft uptime, maintenance and operations planning, and inventory management. Downtime reduction alone provides a strong incentive for better solutions, as every hour of unplanned downtime has a costs of about $10,000.

At the same time, maintenance interventions are expensive. A team of engineers needs to inspect the components affected and conduct an investigation to verify the malfunction, diagnose its causes and finally define and implement a solution. 
For this reason, it’s important to limit the number of false alarms.

Finally, there is a data availability constraint. New aircrafts undergo a limited number of test flights before being released and an anomaly detection system has to work from day one. A dataset of well labelled anomalies and failures simply doesn’t exist.

The solution

The client’s maintenance strategy relies on legacy rule-based methods to detect anomalies paired with time-based inspections and engineering redundancy.

Starting from sensor data of just a handful of test flights, 
Alpha-i is building data-driven health models for two key components of the aircraft.

The ⍺I Predictive Maintenance platform will allow the client to continuously monitor new flight data and catch problems faster, even if they had never happened before.

Confident detections will minimize the number of false positives, while sensor diagnostics will guide the engineering team in investigating the issue.

Business results

This project is still in progress!

Stay tuned too see what business results ⍺I Predictive Maintenance can deliver for our customer!

Project Highlights

  • 6 weeks project
  • Work in progress!

⍺I Predictive Maintenance

Your AI support platform that helps you maximize uptime and prioritize equipment maintenance by integrating with your business processes and autonomously analyzing all your monitoring data.

Product Trials

We provides 6-12 weeks trials to tackle your company’s operational challenges and demonstrate the business value that Alpha-i can add to your organisation.