Introducing Machine Learning
One of the fundamentals of efficient operations is the requirement to optimize asset availability. Leveraging the most out of the investment in your assets is critical, and the best way to do that is to minimize the scheduled and, more importantly, unscheduled downtime. This need makes it easy to understand why the maintenance and predictive analytics market is growing at an annual rate of 22% and projected to be a 24.7 billion market by 2019.
One of the most intriguing products in this market is Mtell, with a unique focus on machine learning and cross machine knowledge sharing. IT Vizion, Inc is a channel and delivery partner with Mtell to provide this capability to our customers – often integrated with our other supported products such as XHQ and OSI PI. This is tied closely to IT Vizion objective of providing B2RT systems solutions.
MTell, is looking to improve on the traditional maintenance approaches which have proven to be ineffective – either providing maintenance downtime that was not needed (per Emerson, 63% of maintenance), or maintenance downtime that does not prevent asset failure (85% of assets end up failing despite routine maintenance).
MTell has creatively extended the typical pattern matching of predictive analytic tools to provide integration into maintenance and/or reliability systems to capture event history to reinforce the actual failure events captured in the signal history in the process historian. IT Vizion can facilitate and improve on this with integration through the Siemens XHQ product – it can provide a data service layer through a custom connection to simply and enhance the data feed requirements of Mtell.
Another unique approach support by Mtell is the notion of shared libraries of fault patterns that can be shared and/or distributed across assets of like type and like function. Not only is this of value within an organization, as faults need to be infrequent and randomly distributed across assets, but the ability for Mtell or IT Vizion to build fault pattern libraries from the varied customer experience will add value for all customers and possible prevent a fault that was never seen before within that facility.
The machine learning aspects not only allow the identification of fault patterns for early warning and prediction, but also the ability to identify the various ranges of normal or non-impactful operating conditions. This allows the software to send warnings when the process starts deviating from the known conditions and allows this new condition to be mapped to a good or to a fault pattern. In the second case, you have used previous history to eliminate failure condition, even though that condition had never occurred!
Mtell has been available since 2006 with their scientists working on the algorithms and approaches for years prior to that. The target markets to date have been oil & gas, mining, and water but the software is really applicable to any manufacturing process or transportation industries.
Interested in a personalized discussion?
