IIoT Still Requires Verification

Contrary to the sensor message, verification indicated that the tires were neither flat nor carrying a pressure of 500 psi.

By Howard W. Penrose, PhD, CMRP, MotorDoc

When considering digital transformation and IIoT applications, common thoughts are that computer and machine-learning systems will fill a shortfall in technical and trade employees. This includes the use of sensing technologies to monitor assets as a replacement or enhancement of condition-based programs. However, field experience shows that an even higher level of expertise is required of personnel in relation to the technology applied and the equipment to which it is applied.

This can be extended to IoT systems we use personally. For example, a few days before writing this article I had an incident with my business vehicle. I have several independently installed IoT security devices and I pay a monthly fee for the car company’s sensor system that will identify theft attempts and monitor maintenance and conditions. I awoke to an alert from the company’s service that four tires were flat. As I was staying in a depressed neighborhood, I immediately thought the worst and quickly dressed and headed out to the parking lot. I then checked the phone alert application and found that the tire inflation was exactly where it was supposed to be but that the system had decided that pressure all four tires had to exceed 500 psi.

I posted a short article on the subject and was immediately accosted by a few people in the automotive industry that the problem had occurred for a large number of people, so it wasn’t a problem. The company had sent a required update when I logged into the app later that day. No explanation was provided, which has reduced my confidence and trust in the system.

This type of situation is more common than we like to realize, but it is carefully not discussed.  The primary concern with critical equipment is whether IoT devices, especially those that self-learn (marketing speak for ML or AI devices), are set to alter system operation or shut it down. In one case, while at a site that was testing sensors, a data scientist came into a meeting to announce that the temperature of a bearing on a critical motor was exceeding 50000 C. Inspection revealed that the sensor had been damaged by a moving piece of equipment and that the bearing was fine. We also helped the company tech understand that, at that temperature, we would have been incinerated.

Modern IIoT systems are still in their infancy, and many are experiments requiring much more time and acceptance before the data and history will be available to trust them explicitly. Even then, their accuracy will not meet the implicit capabilities of a human being and the ability to explore new concepts outside of general training and computerized experience. In effect, experienced personnel who understand the IIoT devices, systems they are evaluating, and the overall context of the operation are still required to verify indicators generated by the IIoT devices. EP

Howard W. Penrose, PhD, CMRP, is president of MotorDoc LLC, Lombard, IL (motordoc.com). He chairs the wind-power standards and government relations participation for American Clean Power (ACP/AWEA), holds various IEEE standards positions, and is a past chair of SMRP. Reach him at info@motordoc.com.