The customer is a major pulp and paper company in Latin America and is one of the largest in its sector in the world, with a presence in 60 countries. After a series of mergers and acquisitions, in 2019 the organization’s annual capacity more than doubled to nearly 12.5M-ton/year.
As part of its sustainable growth strategy, the customer wanted to improve savings and OEE by shifting maintenance processes from reactive to predictive. To that end, customer chose to pilot SKF Enlight AI’s Automated Machine Learning predictive maintenance on a critical asset with a high rate of failure: the pre-bleaching system motor pumps.
The pre-bleaching process is sequential, with pulp traveling through pumps to different machines involved in the chemical preparation for the final bleaching process. Every time a pump fails unexpectedly, the entire pre-bleaching system must be shut down until the failed pump is fixed. Due to recurring unexpected failures, the pumps were becoming a production bottleneck, causing annual production losses of hundreds of thousands of dollars.
Process flow diagram for a section of the pre-bleaching system.
To minimize unscheduled motor pump failures and their mounting impact on maintenance expenses and logistics, the customer wanted to attain greater visibility into pump health and receive earlier warnings of asset failure.
The business problem
- Annual production losses of hundreds of thousands of dollars
- High overtime labour costs
- Reduction in operational efficiency from repeated production stoppages and emergency work order scheduling
The client chose one reference mill to pilot the SKF Enlight AI Predictive Analytics solution. During the trial period, data generated by over 300 sensors was streamed to SKF Enlight AI’s cloud where it was processed by the solution’s advanced Automated Machine Learning engine.
Based on detection of subtle anomalous and indicative behavioural patterns, the solution provided alerts of evolving motor pump failure.
Quality Prediction Rate: 80%
Time to Failure: ~13.5 days
The bottom line
On average, SKF Enlight AI was able to accurately predict 80% of pump failures and provide a time to failure estimation of 13.5 days. These findings demonstrated SKF Enlight AI’s ability to increase availability and improve operational efficiency.
As a result of the initial cost savings estimations of this pilot, the customer has entered negotiations to deploy the solution within 300 assets from 10 different asset families across 3 reference factories.
Once SKF Enlight AI is fully deployed across 7 sites, the organization expects the solution to save US $834,000 annually on the pre-bleaching use case alone.