Next Generation Smart Sensors for the Process Industries: Embedded AI in Process Sensors

KEYWORDS: Smart Sensors, Embedded AI, Process Industries, Machine Learning (ML), Edge AI, Generative AI, Large Language Models (LLMs), Agentic AI, Self-Diagnostics, Microcontroller Unit (MCU), ASIC (Application-Specific Integrated Circuit), Inference

Overview

In the industrial sector today, new technologies such as generative AI, large language models (LLM), agents, agentic AI and edge AI are hotly debated, hyped, and even feared, from the cafeteria to the boardroom. As another form of industrial AI, embedded AI is gaining attention in industrial automation, albeit in a far more discreet and unobtrusive way.

Embedded AI has reached the level of practical application in sensors for the process industries where they will support smarter, autonomous operations. Vendors claims that smart sensors enable more sophisticated real-time analytics than currently possible, and the execution of tasks like self-calibration with or without network connections.

Today’s “Smart” Sensors

While sensors with embedded AI are still in the development phase, smart sensors have already captured a significant share of the market and continue to grow steadily. According to the ARC Smart Sensor research report, the market for smart sensors will grow at 6 percent per year, far outpacing growth in the overall sensor market.

But what exactly is a smart sensor? Does it mean that every smart sensor carries AI functions and can perform advanced data analysis? Not necessarily. In industrial practice, a smart sensor is defined by two key capabilities:

• Self-diagnostics and status reporting – for example, if the lens of a photoelectric sensor becomes contaminated, it can signal the operator directly.
• Remote configuration via electronic interfaces – for instance, when installing a new sensor, the operator does not need to configure it entirely from scratch; settings can simply be transferred from the sensor it replaces.

Why then, do most smart sensors not include built-in functions for advanced diagnostics and analytics? Often the reason lies in the application context. For the majority of “classical” use cases, it is still simpler, faster, and more cost-effective to transmit sensor data upstream for analysis or other tasks to the edge computing nodes like gateways, industrial PCs, SCADA, or other dedicated data platform level, or even directly to the cloud. This approach not only satisfies most industrial requirements but also simplifies fleet management.

ARC Advisory Group clients can view the complete report at the ARC Client Portal.

Contact Us if you would like to speak with the author.

Obtain more ARC In-depth Research Market Analysis.