With the recent advancements in and increased availability of techniques and solutions leveraging machine learning and AI to address various challenges in the industrial sector, there has been uncertainty about the role of vibration analysts in this new era.

Predictronics CTO Dr. David Siegel discussed this issue during the Mobius Connect Global Week of Webinars (WoW) with co-hosts Ed Spence, from The Machine Instrumentation Group, and Matthew Moore, from the Mobius Institute and the Mobius Connect Community.

This webinar not only addressed the strengths and weaknesses of human intelligence and AI, but it also highlighted the opportunities and impact that can be created when humans and AI integrate their unique capabilities to address challenges in industry more accurately, comprehensively, and efficiently.

Below are our takeaways from the webinar, highlighting the pros and cons of the human and AI approaches:

1. Processing Power - Human vs. AI

• Human – While human intuition and experience can provide critical insight for processing data and selecting features, evaluating thousands of raw data points and dozens or hundreds of sensors can be a near impossible task for humans to handle alone through traditional route-based analysis.
• AI – Processing thousands of data points from hundreds of sensors can be tackled with ease by AI, making the jobs of humans easier and ensuring the overall diagnosis process is more cost-effective and efficient. However, there can be a loss of granularity due to the global approach of data processing in the diagnosis stage.

2. Method – Human vs. AI

• Human – With the help of vibration analysis software, humans use past data to learn normal machine behavior and identify patterns and trends in similar machines, making decisions on an individual basis. While this approach can yield accurate conclusions for specific assets, it is a challenge to apply at scale and in real-time, which can be important in modern industrial applications.
• AI – Data availability and data quality checks, along with relevant baseline data for early detection from past failure cases, can be leveraged in the development and training of accurate AI models and can allow them to be deployed at scale. However, flawed data inputs lead to flawed results and insights, i.e. garbage in, garbage out.

3. Accuracy and Value – Human vs. AI

• Human – Asset failures can be overlooked by humans simply due to the sheer quantity of data that must be reviewed. An analyst may require extensive educational training and experience in the field to identify all issues and make consistently accurate decisions. Humans often need the second opinion of a colleague before making the “right call.” However, humans continuously learn over time and use this knowledge to make decisions based on past experience, learning from their mistakes.
• AI – AI can analyze large quantities of data with great accuracy and provide even more value and robustness over time when trained and fine-tuned by humans with expertise. It should be noted that AI alone does not learn continuously, meaning these technologies may take decades to mature and, even then, AI may never be able to truly think outside the box for itself.

4. Training – Human vs. AI

• Human – Human experience and expertise are invaluable; however, proper certifications and years of experience are needed by humans in order to accurately compare similar machines to determine a normal baseline vs. impending failure, detect patterns and trends of specific fault frequencies, and master vibration analysis.
• AI – A human-in-the-loop who has knowledge of AI algorithms, as well as specific domain knowledge and expertise within the target industry, is needed to train and fine-tune AI models for AI to accurately predict, and diagnose, issues and failures.

5. The Future of Vibration Analysis – Human vs. AI

• Human - Vibration analyst jobs are safe and AI is not “taking over.” A human-in-the-loop will always be needed in order to evaluate, confirm, and fine-tune the results obtained by AI models.
• AI – AI takes on the menial task of sifting through countless data points to identify trends and patterns, opening up humans to more complex, creative, and innovative tasks.

While there are many pros and cons to the individual approaches of humans and AI, an experienced vibration analyst can continue to use their domain knowledge to work with accurate AI algorithms as a human-in-the-loop, helping to train and understand, in order to best inform asset decisions for their company.

For more examples on AI solutions that fit into this paradigm, check out Predictronics’ past webinars on automotive manufacturing and semiconductor manufacturing applications, among others.

Week of Webinars Global, organized by Mobius Connect, is a week-long series of educational presentations delivered by industry experts that focuses on condition monitoring and reliability improvement. Mobius Connect provides professionals with reliability and CBM resources and hosts a community for leaders and practitioners to share challenges and solve problems.