IoT Predictive Maintenance

One of the biggest challenges in large-scale industrial machine deployments today is to keep track of the functioning of the machines, and to minimize the downtime. There is a huge benefit in being able to predict accurately the failure of a particular machine.

At Persistent, we understand this challenge and have build solutions that:

  • Integrates IoT sensors data from diverse sources.
  • Profiles the sensory data, to detect anomalies and raise alerts.
  • Predicts future breakdowns and minimizes operative costs.
  • Promotes ideal usage of machines by detecting incorrect settings.
Customer Success Snapshot

A specialist in design, manufacture and support of electrical control panels, displays, remote controls, and wiring systems for all engine based machines.

Challenges
  • Accurately predicting a machine’s future breakdowns was difficult that resulted into high operation costs, increased downtime, and delayed repairs. This called for integration of machines to send real time sensory data.
  • Each machine had a different working profile which had to be learnt based on historical data before anomalies could be identified. Using a manual which gave approximate running parameters was not sufficient to get the working profile.
  • Data from sensors was often missing which made learning the machine profile difficult.
Persistent Solution
  • Developed an end-to-end solution integrating data from sensors deployed on mechanical machines in the field and analyzing the data at run-time to detect anomalies.
  • Historical data (cleaned up using ETL processes) was used to develop a Machine Learning system to profile the engines and learn their default behavior in the presence of other seasonal factors.
  • Following conditions and parameters were tracked for an alert-based dashboard system:
    • Determining fuel economy degradation to alert if more fuel would be used for similar workloads.
    • A probabilistic model to learn the ideal oil pressure value at different engine speeds, to raise alerts for system diversions from these values for a prolonged period of time.
    • To alert when system observed an abnormal amount of time taken for the oil pressure to stabilize at engine start-up time.
Result
  • Reduced number of breakdowns, raised alerts when the system behaved abnormally, and reduced downtime due to timely preventive maintenance of machines.
  • Reduced overall cost of operation by optimizing operational parameters and reducing human oversight.
  • Helped to study the wear and tear in machines whenever the new profiles diverted from the ideal working profiles.
  • Inefficient usage by operators was detected and avoided which resulted in considerable amount of fuel savings, and helped prevent wear and tear.

You May Be Interested In

VIEWS

Whitepaper: Facial Analysis: Face Recognition, Expression Recognition, and Gender Identification

VIEWS

How machine learning can help with demonetisation and counterfeit currency

VIEWS

Anomaly Detection in IIoT: A Case Study using Machine Learning

VIEWS

Machine learning will be the most exciting tech trend of 2018: Persistent Systems CTO

OFFERINGS

Engage Persistent Systems to conduct a Machine Learning Workshop for you!

OFFERINGS

Watch how ChatOps uses AI & ML to integrate and automate DevOps process

Start typing and press Enter to search

Contact Us
close slider

Contact Us
Have a question? Drop us a line and we will get in touch with you!

Yes, I would like Persistent to contact me on the information provided above. Click Here to read our full Privacy Notice.