Can ChatGPT replace the Reliability Engineer?

If you have been hanging around the Internet lately you have probably heard about OpenAI’s platform which recently had one of their products, ChatGPT go viral after its launch in November 2022.  Microsoft is also in discussions with OpenAI to potentially invest $10 Billion dollars into the company in addition to their initial $1 Billion investment.  That was not a typo, $10 Billion dollars in new investment with the intention to integrate the technology into the Office product suite as well as their search engine Bing.

ChatGPT is a free AI application (Neural Network based) which has scoured the internet for information and can answer questions in a very ‘human-like’ fashion.  The capability is very expansive, some examples we experimented with included defining job descriptions, writing essays on English literature, solving bio medical engineering equations, and even writing Python code!

Our experimentation with ChatGPT got the Itus team wondering, if we set asset operating context aside, could it define an asset strategy?

To satisfy our curiosity, we decided to run some scenarios and see what we got back.  Our first request to ChatGPT was to ask it to define the failure modes for a specific asset class.  On this task we would say the results were a bit better than expected as we received a relatively good list of failure modes like what someone with a reliability background would come up with.  Asking it about a centrifugal pump it returned the following:

Common failure modes for a centrifugal pump include:

• Impeller wear or damage
• Shaft or bearing failure
• Clogging or blockage of the suction or discharge lines
• Cavitation (vaporization of the fluid due to low pressure)
• Worn or damaged mechanical seals
• Loss of prime (loss of suction)
• Air binding (air getting into the pump)
• Overheating
• Electrical or control issues.

 

Then looking specifically at impeller wear we inquired about mitigating maintenance or condition monitoring which could be implemented, to which we received the following:

Some maintenance and condition monitoring activities that can help mitigate impeller wear in a centrifugal pump include:

• Regularly inspecting and cleaning the pump’s impeller and other components to ensure they are free of debris or other blockages.
• Monitoring the pump’s vibration levels and bearing temperatures to detect any abnormal conditions that could indicate wear or damage.
• Checking the alignment of the pump’s shaft and impeller to ensure they are properly aligned and not causing undue wear.
• Measuring the pump’s flow rate and head pressure to ensure it is operating within its design parameters.
• Performing regular lubrication and oil changes, as well as checking the oil’s viscosity and contamination, this help to maintain the pump’s bearings and mechanical seals.
• Checking the pump’s suction and discharge lines for blockages or restrictions.
• Implementing a preventative maintenance schedule and following the pump’s manufacturer’s recommendations for regular maintenance.
• Conducting regular ultrasonic testing of the impeller to detect any defects.
• Inspecting the impeller’s wear ring and replacing it if it worn or damaged.

 

Not a bad list but it is generic and lacking details.

We explored further and asked for an FMEA on the same asset class.  The result was mixed.  At first, we were impressed with how it organized the results by typical FMEA definition and included potential causes, potential severity, risk priority number, and recommended actions in the results.   Optimism faded when we realized we only got two failure modes back.  On subsequent attempts, we were more specific with the asset class and service of the fictitious asset.  The results changed very little.

So where does this leave us?  Does ChatGPT have a role in a reliability department?  Could it be utilized to enhance strategy development approaches such as RCM or FMEA?

If you had nothing, ChatGPT could be used to get generic information about failure modes, failure rates and PMs to mitigate those failure modes.  While this is more helpful than an empty document or spreadsheet, it still leaves the real effort of implementing the asset strategy as defined by the FMEA untouched.

This is where the Asset Performance Management (APM) process can be an enabler.  APM is rooted in the methodologies of RCM and FMEA and implements their output in the form of the asset strategy.  Modern APM solutions offer built-in library of strategies which will be much more useful than what ChatGPT can currently provide.  However, a library is not enough.  The value of APM is only realized when the strategy is implemented and “operationalized”.  This is where modern APM solutions drive tangible value for industrial organizations seeking to maximize availability, optimize maintenance spend and lower risk to people and the environment.

If you are a student looking to write an essay on The Lord of the Flies, you could certainly consider ChatGPT for some insights.  However, for the maintenance and reliability domain, we recommend you tap into the equipment and process experts in your organization, the actual operating history of your equipment and the expertise built into APM applications which will implement asset strategies inclusive of PMs, analytics, monitoring and prescriptive actions.

We certainly look forward to seeing this class of technology grow and what it can do and the creative ways it will be applied.  It will also be interesting to see what it learns from this article which is now publicly available for its engine to analyze!

At this point we can easily state – the reliability engineer role is going to be around for a while!