Identifying and Managing Risks of Sensor Drift

And how machine learning can help to solve it at scale.

Identifying and Managing Risks of Sensor Drift

Identifying and Managing Risks of Sensor Drift 560 420 APERIO

Sensor drift is a common issue in the industrial world that can have significant impacts on the accuracy and reliability of sensor data. It refers to the gradual, subtle changes in the sensor that happen over time, causing a discrepancy between the physical state that is being measured and the output of the sensor.

There are several factors that can cause sensor drift, including environmental conditions, wear-and-tear, and manufacturing faults in the sensor itself. Ongoing exposure to extreme temperatures, humidity and pressure  can cause a shift in the readings of a sensor that are often imperceptible to operators.

Wear-and-tear is particularly common in sensors that are exposed to physical stress, such as vibration, shock, and other mechanical forces. In these cases, mechanical forces can cause the components to move or change, leading to a gradual shift in the readings.

The Risks of Sensor Drift

The biggest risk associated with sensor drift is operators making bad decisions based on incorrect sensor data.

Imagine this scenario: a chemical plant that produces a flammable substance, relies on a network of sensors to monitor various aspects of the production process, including temperature, pressure, and chemical composition. One of the critical sensors in the plant is one that measures the temperature of a reaction vessel where the flammable substance is produced. The temperature needs to be closely controlled to ensure the reaction proceeds as intended and doesn’t become too hot, which can cause an explosion.

Over time, the temperature sensor begins to drift, and its output becomes increasingly inaccurate. The plant’s control system, which relies on the temperature measurement to regulate the temperature of the reaction vessel, fails to detect the drift and continues to rely on the sensor’s faulty readings.

As a result, the temperature of the reaction vessel gradually rises above the safe threshold, eventually leading to an explosion. At this point, the best-case scenario would be an automatic plant shutdown, and a complete evacuation of the plant. The shutdown and the subsequent investigation into the cause of the explosion would result in significant production losses, financial losses, and damage to the plant’s reputation.

Mitigating the Risks

In order to minimise the risks associated with sensor drift, it is important to regularly monitor and calibrate sensors to ensure that operators are getting an accurate reading. In some cases, this may involve regular replacement of the sensors, particularly if they are exposed to extreme environmental conditions or mechanical stress.

Up until now, sensor maintenance regimes have often relied on the experience of operators and vendor guidelines. However, the rapid growth in the number of sensors used in today’s modern plants, has made the identification of sensor drift an increasingly complex challenge. In addition, when operators use multiple sensors to measure the same asset for increased safety and redundancy, there is often no reliable way to discern the good sensors from the bad and a full audit of all the sensors becomes necessary. Obviously, this is an inefficient way to go about things.

An illustration of how sensor drift can occur when multiple sensors are measuring the same asset, without operators noticing it.

Luckily machine learning tools such as APERIO DataWise™ can be used to identify sensor drift at in a reliable and scalable way, even when analysing hundreds of thousands of sensors in a single site and with multiple sensors monitoring the same assets.

By training algorithms on historical sensor data, machine learning models can ‘learn’ the normal behaviour of the sensors and detect anomalies in real-time, even anomalies that may not be apparent to humans, such as gradual shifts in sensor output over time. Once an anomaly or drift is detected, APERIO’s machine learning models can trigger alerts that enable operators to take timely, corrective actions to prevent equipment failure or safety incidents.

Tools such as APERIO DataWise™ can also optimize maintenance schedules by predicting when sensors are likely to drift or fail based on historical data. By leveraging machine learning to identify sensor drift at scale, companies can improve safety, reduce downtime, and increase efficiency in industrial processes.

To to learn more about APERIO’s category-leading data quality solution, book a demo with one of our team or download our data quality E-book:

    Privacy Preferences

    When you visit our website, it may store information through your browser from specific services, usually in the form of cookies. Here you can change your Privacy preferences. It is worth noting that blocking some types of cookies may impact your experience on our website and the services we are able to offer.

    Click to enable/disable Google Analytics tracking code.
    Click to enable/disable Google Fonts.
    Click to enable/disable Google Maps.
    Click to enable/disable video embeds.
    Our website uses cookies, mainly from 3rd party services. Define your Privacy Preferences and/or agree to our use of cookies.