The state of automation is rapidly changing, and one of the main reasons is smart devices. With such capabilities as embedded computing and storage, wired and/or wireless communications, the ability to interact with the outside world or perform autonomous actions, or some degree of descriptive or predictive analytics, smart sensors are adding value to automation technology.
In one example, a smart gas chromatograph has a built-in “software assistant” that guides even less experienced personnel through most installation, operation and maintenance procedures.
Many conventional sensors now also have additional smart features such as pressure sensors that detect electrical loop issues, temperature sensors that can detect thermocouple degradation, and radar level gauges that include self-calibration capabilities. Video cameras can now have onboard video analytics for applications such as intrusion detection. Vibration sensors can perform fast Fourier transform (FFT) of the data to compute equipment health by analyzing the various vibration frequencies right on the sensor. Smart sensors can be attached to the electrical line of industrial electrical motors to compute power factors and the probability of failure for key failure modes—all from analysis performed right on the sensor.
Smart sensors represent a natural technology evolution from simple, equipment-mounted mechanical indicating sensors to “dumb” pneumatic and analog electronic sensors capable of transmitting raw measurements to another device for “massaging” and transforming to today’s microprocessor-enabled smart field devices with onboard processing capabilities and often full, bidirectional digital communications.
Today’s smart sensors can generate 20-50 times more readings beyond the primary value (PV). Some of these secondary values (SVs) are associated with statistics such as standard deviation of the PV, mean of the PV, and maximum and minimum PV values. Other readings are associated with sensor health; these include sensor drift, sensor degradation, and calibration verification.
Rather than sending raw, unprocessed data, today’s smart sensors also produce more refined readings—for example, an ultrasonic corrosion detection tool that sends wall thickness, rather than the raw ultrasonic transducer (UT) waveforms. Business rules are now stored in the device rather in a centralized system, allowing the smart device to perform autonomous actions based on those business rules (e.g., turn the pump off if the tank level reaches a certain level). As devices get smarter, they might not even transmit the raw data; just the results of the analytics performed on those data. For example, a hyperspectral camera might transmit gas levels rather than just the video images, which would require further analysis.
All this means that smart devices are uniquely qualified to be key enablers for the Industrial Internet of Things (IIoT). When you consider the key elements of the IIoT—sensors, networks, Big Data, analytics and visualization—smart sensors contribute to all these areas.
Distributed acoustic sensing provides a good example. Here, a fiber cable is inserted into a producing oil or gas well, and onboard processing transforms light pulses into a measure of acoustic energy at specific points about every meter along the length of the cable. These measurements are then used to optimize production by improving visibility into the reservoir behaviors. Since these systems can acquire over a terabyte of data per day, this new sensor is bundled into a smart sensor package that provides methods for networking, handling Big Data, computing analytics, and visualizing the data.
The introduction of these smart field devices poses some significant challenges for technology users. With so much intelligence being pushed to the field device, end users need to distribute, verify and audit those rules in a distributed environment. Though reasonably straightforward to do in a centralized system, managing those business rules in a non-centralized world is far more challenging.
The networks for gathering and backhauling readings must also be designed to accommodate all the secondary readings being generated. The classical ISA-95 security model in which all the data are funneled through the DCS and process control network does not scale easily. Instead, an enterprise approach is needed to avoid overwhelming the process control system with this tidal wave of new non-control data.
Another key issue is that while the smart devices now generate and communicate sensor health readings, an enterprise device management solution is needed that can collect and analyze this sensor health data. Such systems could significantly improve the reliability of these smart sensors.
Finally, systems need to be designed to collect, process and transform all this additional secondary, non-control data into actionable information to help improve the performance of the plants and associated assets. Traditional systems greatly underutilize such data.