Anomalies, False Positives & False negatives

In this podcast, Sunil – Founder & CEO, and Graham – VP, Business Development, talk about Anomalies, False Positives &False negatives, and how they affect Machine Health Monitoring and Predictive Maintenance.

The podcast begins with Sunil introducing Nanoprecise, and how the combination of our hardware and software offers an end-to-end solution that prevents unplanned downtime. He points out that various factors such as power source for hardware, connectivity etc. impacts the data quality and the accuracy of the results, thereby limiting the opportunity to scale, and that resolving these issues have helped Nanoprecise to better predict failures.

Sunil explains that an anomaly can be a fault or a process upset. It can be cavitation or a variation that has nothing to with any component going bad. He details that for any software, anomaly detection will be higher in the beginning, and it will come down over a period of time as the software learns to differentiate between process upsets and faults. “A good software will differentiate between anomaly/fault severity and offer customized methodologies to treat it.” says Sunil.

Graham and Sunil talk about using Artificial Intelligence for reducing false positives and false negatives. They agree that data is an integral component for AI & Machine Learning algorithms, and that the Output Parameter Data of each machine acts as a signature, that can be used to co-relate their spectrum with corresponding faults. They point out that Nanoprecise uses AI + Physics based models, which uses multiple sensing elements & modalities to screen process-related parameters, and identify the faults based on their frequencies and patterns.

Sunil also provides the example of a unique application for our automated AI-based Predictive Maintenance System in Light Rapid Transit and the learning’s that have helped us improve the implementation of our solution in machines with variable speed.

The podcast was concluded by agreeing that it is essential to set realistic expectations from maintenance systems, by clearly defining anomaly, false positive and false negative. This is because contradictory ideas of these terms in the minds of operators and service providers can result in pilot purgatories.

Tune in and learn what false positives and negatives are and how they will affect your monitoring program.