How do manufacturing and other industrial organisations move from traditional, scheduled maintenance to proactive, predictive maintenance? What are the first steps required to begin this transformation? What are the different stages in the journey, from the initial business case and design stage, through to data readiness, cultural adoption, and scaling?
In this film, Senseye's Alexander Hill and Rob Russell are joined by Peter Gagg from MCP Consulting Group to explore how predictive maintenance isn’t simply a technical challenge with a technological solution, but an ongoing journey involving knowledge acquisition, cultural acceptance, and ever greater collaboration between humans and smart systems.
Alexander Hill, Senseye: Redesigning a maintenance process is a difficult thing, we’d never advocate that you scrap your existing maintenance practices and jump straight into PdM. There’s the old saying that, over-maintenance is always cheaper than failure, and that’s correct.
A balance needs to be found between your scheduled maintenance and your predictive maintenance. Then, gradually, as you trust your maintenance system more and more, you can reduce some of the preventive measures that are expensive but are known to help.
You have to enact this cultural maintenance change quite slowly, as you build trust in the system and trust in the business outcomes of saving money and increasing efficiency.
Rob Russell, Senseye: Maintenance is considered to be built up in layers of maturity, from reactive to planned, to preventative and proactive types of maintenance. One shift that I really think has to happen, is not considering these as stages to go through but selecting the right type of maintenance approach for the right asset.
Peter Gagg, MCP Consulting Group: Predictive maintenance is not the solution to all problems within an organization. It’s got to be used in the right places at the right time.
The use of condition monitoring tools to replace planned maintenance tasks with predictive maintenance tasks is important. However, to get to that position - where you know which pieces of equipment will benefit from PdM - you need to apply techniques such as FMECA (Failure Mode, Effects & Criticality Analysis) and RCM (Reliability-Centered Maintenance) to determine the right combination of maintenance tasks: whether that's predictive maintenance, preventative maintenance, run to failure, or even designing out failures that are occurring.
Rob Russell: We actually think about this in terms of design, deploy, operate, and then refine. The initial steps within design relates to that asset selection but then also having an appreciation of the various failure modes that you have to capture. That then informs the decisions about the correct type of data and information that you need.
That’s all information that then can pass through into the deploy phase, where you can work with your controls engineers and IT teams to implement that data capture and enable it to be streamed to the cloud.
Alexander Hill: We can then start to feed that data into Senseye, that’s an automated process, it will take this data and learn from it and start to build a steady-state of what these machines and assets look like and we can then predict against that.
Rob Russell: The use of AI is fundamental to what we do. It enables us to do the analytics at large scale and enables us to detect things within the data that might be too subtle for a human to pick up. What is really key and important is that it’s done in a human-centric way that delivers the information in a way that is accessible to the end-users and can also become part of their day-to-day work processes.