Last week, I appeared on the “Ask the Experts” panel webinar with my friends at UpKeep – you can listen to the recording here. The first question to the panel was about the advantages and disadvantages of using AI for predictive maintenance.
As some of you know, in a previous role, I learned machine learning online at EdX to analyze oil analysis results. Even though I’m no data science expert, I was able to produce a ~50% reduction in samples that my team would need to look at. Over that project and several podcasts with AI experts from Quartic.ai, Uptake, Petasense and SymphonyAI (see below for the links to all those podcasts), I’ve put together 3 benefits for you.
- Reduce Labor – If you have a repetitive task, you can use AI as a digital assistant to reduce your workload. Something like looking at PdM data (oil analysis, sensor readings, vibration analysis, etc.) is a great task for AI and you can save your human experts for the problem samples.
- Don’t Have to Just Rely on Numerical Data – Compared to using rule-based systems, AI has more flexibility in terms of what data can be analyzed. Stanford University did an experiment using AI to read chest X-rays and it produced great results. Visual inputs has many uses in industry that doesn’t only limit you to predictive maintenance (although you could use it on infrared camera pictures).
- Use it On Processes – Lastly, don’t limit yourself to predictive maintenance! In a podcast I did with Don Doan from Symphony AI, he mentioned that they’ve seen an increase in 1-2% in production by using it to optimize how the plant is being operated. A 1% increase in production is likely worth more than saving labor on your PdM program!
Stay tuned for next week’s blog on the mistakes that people make using AI.
Have you used AI at your facility? If so, hit reply and let me know what you used it for and what the results were. I’d love to hear your story!
Reliability Never Sleeps,
Rob
Artificial Intelligence Podcasts on Rob’s Reliability Project Podcast:
- Using AI to Optimize Processes with Don Doan
- Implementing AI with Adam McElhinney & Derek Valine
- Listener Questions about AI with Rajiv Anand
- Artificial Intelligence Primer with Adam McElhinney
- Machine Condition with Arun Santhebennur
- Merging AI with Failure Modes with Mark Benak
- E4 – Blair Fraser – Artificial Intelligence & Machine Learning
Do you have information about AI implanted to engineering designing? I mean, are there previous experience using AI to improve design or engineering process of process control systems? Thanks in advance
Hi Wilmer, I don’t have any experience with using it but with a Google search, it looks like there are some books about it. Hopefully you can find what you’re looking for!
Great ideas Robert. Can you point me to some good resources on how to use AI/machine learning to evaluate reliability data such as some key performance parameter on units coming in daily and making disposition decisions for good/bad/needs investigation/root cause drill down.