Role of Data in Reliability
With the advent of Industry 4.0, data and connectivity have enabled equipment to become continuous factories of data. Whether process data, master data or realtime data, it has never been a more challenging time to adjust and harness the power of seemingly huge piles of potential information. Traditional maintenance and reliability are undergoing a tremendous shift as organizations become more data-driven. Sean Rosier and Nathanael Ince of PinnacleART are on the show to help us put in context, the relevance of data in reliability.
In this episode, you’ll learn:
- Where to start implementing a data-driven program
- The influence of technologies like machine learning and digital twins
- Getting started on using data for reliability
- Arising challenges and possible solutions around data-driven strategies
Where to start on a data-driven program
Before diving into collecting data around your facility, you have to be fairly clear on the end goal. What state would you like the data to help you reach? Is it knowing your utility consumption vs production metrics? Is it equipment health? Aside from the expected outcomes supported by this data, it is also important to understand the frequency interval and the most crucial data to focus on.
Other considerations can be :
- Plan out a data collection strategy that is robust
- Have a master asset list – Which can get assembled by doing a walk down in the plant, or using P&ID documents
- Have a way to visualize the assets in your portfolio – Which might help in understanding relationships between processes and equipment
What is a digital twin?
A digital twin is simply a virtual representation of a physical asset. It allows us to run simulations on the virtual copies, look at the outcomes and use the results to influence the existing asset. This concept has a direct impact on reliability in that the relevant personnel can run criticality analyses on the virtual models before implementing anything on the ground. Not only does it save time, but it also avoids most of the unforeseen implementation costs. What’s more, reliability engineers can incorporate dynamic (real-world) influences to have an accurate picture of the outcome.
Where do we start with data-driven reliability?
A good starting point would be having the end in mind. This sets the context for the resource requirement in implementing data-driven strategies.
Other steps to consider are not limited to:
- Ensuring you have the foundational data e.g. criticality analyses of assets
- Selecting the appropriate data that will help support your analysis of progress
ML Big Data in Reliability
Smaller companies are putting together innovation groups which proceed to deploy Machine Learning Projects as pilots. However, these pilots are run in isolation from the input of maintenance and reliability personnel. Another challenge is that even with successful pilots, organizations will still have a hard time integrating the Machine Learning models into their operations.
Another consideration is to make sure you identify the problem even before looking for any machine learning solution. A clear problem statement will help guide teams towards which ML strategy to adapt. It should also factor in both intrinsic and extrinsic factors that can then be used as scenarios when training the algorithms.
Challenges facing data-driven reliability
One of the key challenges involves having people on board and with a good understanding of its significance. Often the front line technicians who interact with assets on a day-to-day basis are neglected in formulating data strategies. This challenge curtails an important source of qualitative data input. Therefore engaging and listening to these people on the ground helps in getting a holistic picture of the data strategy.
Without buy-in from leaders in that organization, it becomes hard to implement new data technologies. Instead of running pilots with a single department, organizations should consider involving relevant departments that will be end-users of the technology. The communication and buy-in should also come in as early in the project as possible.
Results of ML on Reliability
From some of the successful deployments of ML, there is usually an improvement in the prediction mechanism. Equipment abnormalities can be detected and the necessary corrective measures are taken well in advance.
Machine learning also enables a new way of looking at assets because of the possible relationships that can be drawn from the collected data. This new approach can help visualize the health of an equipment a lot better.
Challenges of data in reliability
Culture stands as one of the major barriers to adopting these technology tools. Leaders have to recognize that a cultural change precedes a technology change for success to be possible.
Having buy-in from both formal and non-formal leaders in the organization provides good momentum to have the support of the facility personnel. The people who are also championing of buy-in should also be committed to seeing the project through.
One thing to implement
A major takeaway from this episode is that it is paramount to communicate the value of this shift from the throughout the organizational structure. Having a committed party also helps to maintain momentum throughout the project implementation.
Sean Rosier and Nathanael Ince Links:
Rooted In Reliability podcast is a proud member of Reliability.fm network. We encourage you to please rate and review this podcast on iTunes and Stitcher. It ensures the podcast stays relevant and is easy to find by like-minded professionals. It is only with your ratings and reviews that the Rooted In Reliability podcast can continue to grow. Thank you for providing the small but critical support for the Rooted In Reliability podcast!