AI Is Not a Job-Killer

Jane Alexander | April 25, 2019

Advanced technologies can expand your value and make your work more enjoyable.

Artificial Intelligence (AI) is a hot topic these days. Discussions abound about how it and/or the Industrial Internet of Things (IIoT) will change workplaces. Moreover, countless operations are beginning to demonstrate real results from putting AI into practice. At the same time, the situation can be intimidating for maintenance technicians and engineers who have already seen a shift in manufacturing. That’s not surprising.

According to the U. S. Bureau of Labor and Statistics, Washington (, the United States lost 5 million jobs in the manufacturing sector between 2000 and 2015, even as production increased. Some people now worry that AI, IIoT, and other technologies related to digital transformation mean the next wave of job losses is coming. Are their fears justified? 

Definitely not, according to Abhinav Khushraj, co-founder and CEO of Petasense, San Jose, CA ( 

As Khushraj notes, the fact is most of those technological advancements will be deployed on jobs that aren’t being done today. “For example,” he said, “plant personnel can’t simultaneously look at hundreds of process variables on a second-by-second basis to identify small deviations from normal performance. AI can.”


John McCarthy, one of the founders of AI, defined it as “the science and engineering of making intelligent machines.” Khushraj expanded on that definition with the following points:

• Machine learning (ML) is a subset of AI in which the computer has the ability to learn and improve without being explicitly programmed.

• Anomaly detection and classification are two of the most practical applications of machine-learning techniques used in predictive maintenance (PdM) today.

• Anomaly detection is, at its core, classification, except that one of the classes is considered to be an anomaly and the other is considered normal. In PdM, anomaly detection could be used to determine whether the operating condition of a critical machine has deviated from what was considered normal in the past. The baseline, or normal, performance can be characterized by observed values of multiple sensor parameters. For example, for a pump, you could simultaneously measure suction and discharge pressure, flow-rate, vibration (in all three directions) on the drive-end bearing, temperature of the bearing, and current draw on the motor. Through the collection of thousands of such measurements, anomaly detection is able to easily identify any among them that are abnormal. 

• Classification can be used to identify a specific defect and provide a diagnosis. For example, classification algorithms can go beyond simply telling personnel that there is an anomaly to actually predicting cavitation in a pump.

For classification to work, you need to acquire a body of historical performance data, which can require considerable time and money. In the case of cavitation, this information would be identified as data from actual instances of pump cavitation and then used to train the predictive model to recognize conditions that signal future cavitation incidents..


Khushraj acknowledges that anomaly detection and classification aren’t new concepts. “Vibration-analysis software,” he explained, “had rules-based expert systems to codify knowledge and diagnose problems 15 to 20 years ago.” But rules-based systems have several issues: They’re painstaking to develop. They’re only as good as the few people codifying the rules, and as, Khushraj put it, “They’re really hard to maintain, especially when presented with new failure patterns. As a result, most plants didn’t implement rules-based systems. Moreover, not enough data was collected at a scale to truly test their efficacy.”

Machine-learning-based systems eliminate such problems by minimizing the need for extensive human intervention. Unsupervised anomaly detection doesn’t require any human intervention. Even supervised classification systems only need an adequate amount of labeled data upfront.

“Once you get this data,” Khushraj said, “you can re-purpose the models across a wide variety of use-cases.”

As for the question of “why (AI) now” Khushraj pointed to three trends driving the explosion of interest in machine learning for PdM: 

• Decreased sensor costs means it is possible to equip more assets with permanent sensors and collect substantial data at scale.

• Cloud technology makes it affordable to securely process, collect, and store data.

• Machine learning can be applied to make sense of all data that’s being generated.


Technology by itself won’t change anything if there isn’t support in implementing it. Success or failure of any initiative depends on the human dimension, so it is important to understand the perceptions of people who will be working with it. 

Khushraj said his organization has found that the three main reasons people are reluctant to try ML-based technology are all based on the human element. The most common reason is associated with the possibility of the system missing failures. “There’s been a lot of hype around ML for PdM,” he explained. “And people are skeptical of promises that ML will solve everything. As a result, it can be difficult to tell what is reality, and no one wants to put his or her job on the line for technology that isn’t going to work. “

Another common fear is that of “we will be replaced by computers.” In manufacturing today, anything that seems like it could result in headcount reduction is sensitive. This often is expressed in statements along the lines of, “We’re already doing PdM using walkarounds and it’s working for us.” According to Khushraj, that approach may, in fact, be working well, but there could still be opportunities to improve through more frequent data collection or automated analysis.

A third reason people are reluctant to implement ML-based systems for PdM involves the required skills. Most plants don’t have data scientists on staff, and the hurdle to retrain for these jobs is large. 


Today, most plants are conducting PdM by using in-house personnel or third-party service providers to collect data on a monthly or quarterly basis. Khushraj points out that, whether done in-house or outsourced, the interval between readings can result in failures that develop more quickly than the intervals. “The falling cost of wireless sensors,” he said, can help to address this by providing more-frequent data.”

But, Khushraj continued, more-frequent data also means more human analysis, which often isn’t available. He described the experience of electric utility Arizona Public Service (APS), Phoenix (, which implemented a wireless PdM program and, within the first six months, collected more than 1.4-million readings and caught 13 defects. In comparison, APS’ old walkaround program would have only collected 5,000 readings, most of which wouldn’t have corresponded with the key operating hours.

With ML, millions of data points can be run through advanced algorithms that are constantly improving and creating better prediction models. It would be physically impossible for a human analyst to process the quantity of data that AI, ML, and IIoT advancements create. “Of course,” Khushraj emphasized, “more data doesn’t mean more insights unless something is done with it.” 


According to Khushraj, the best PdM programs view ML as an assistant to in-house reliability staff. Rather than replacing engineers or technicians, it enables them to focus only on the assets that need attention, obtaining more data to make better diagnoses, with real-time alerts. 

“ML,” he said, “is perfectly suited for the mundane tasks that humans aren’t suited for, as in finding patterns in millions of lines of data. Humans are far superior when it comes to determining the meaning of the data.”

Experienced reliability personnel can add their knowledge to train the models on failure modes. Instead of the system only detecting an anomaly, it could alert personnel of specific problems. “Today,”  Khushraj stated, “there are still no systems that can accurately tell you all of this information. The knowledge of the experienced individual is critical in understanding what to do about the potential problem. In short, humans need ML and ML needs humans.”


With routine data analyses increasingly performed by algorithms, Khushraj said reliability personnel can apply their knowledge to preventing failures. As an example, he described a recurring pump failure due to undersizing for the application. “The underlying root cause or ‘why,’” he said, “can’t be captured by an algorithm.”

Eliminating mundane tasks for certain personnel has other advantages. Among them, it can help shift plant maintenance efforts from reactive to predictive to prescriptive. Similarly, an ML-based system can take into account parameters that have not traditionally been monitored, such as simultaneously examining operations and condition data. 


Instead of completely overhauling their existing PdM programs, he encourages personnel to look for areas that can be improved, i.e., assets that are not currently being monitored or ones where failure is happening in between routes. Use the algorithms in tandem with existing analyst knowledge so you can validate that it is catching the same failures a human would catch, without giving false positives. This will help to reassure everyone that it will work. 

“If you don’t have data scientists on staff,” he advised, “find a provider offering pre-built applications that don’t require specialist skills.” Keep in mind, though, that a general-purpose ML system capable of building algorithms for any application, i.e., operations, maintenance, finance, will require more data-science knowledge than one built for a specific problem. 

“Identify the problem you are trying to solve,” Khushraj said, “and select a few companies to pilot, testing the algorithms against each other to see which one performs the best, and is the easiest to configure and use.”


The bottom line for Khushraj is that change is going to come, whether we like it or not. Organizations (and workforces) that fear advanced technologies (or are slow to adopt them) will be replaced by others that view technology as a tool to help improve human efficiency. “Embrace the opportunity of having an assistant that can help sort through repetitive tasks, allowing you to do the things you love,” he urged.  “The smarter the assistant, the better you will be at your job.” EP

Petasense ( offers an end-to-end solution with a wireless vibration sensor, cloud software, and analytics that supports asset reliability and PdM. 



Jane Alexander

Sign up for insights, trends, & developments in
  • Machinery Solutions
  • Maintenance & Reliability Solutions
  • Energy Efficiency
Return to top