What is a false positive?

A false positive, or type 1 error, is a result which indicates that a certain condition is present when it actually is not. In contrast, a true positive is when the results correctly show that a given condition is present. A false positive rate means the percentage of total alerts or test results which are falsely positive.


In manufacturing, this could be a false alert, which warns the control room about a problem in the system even though the problem does not exist. It could also be a false result, for example, if a piece of equipment is tested, and the test concludes that it is not functioning properly and needs to be fixed, even though the item is working as it should.


It’s also possible to receive a false negative result, which fails to pick up on a condition or anomaly and instead indicates that everything is working as it should. Type 1 errors can affect many situations, including medical tests and cybersecurity systems as well as process manufacturing plants.

Why does a false alarm matter to process manufacturing plants?

It’s common for engineers and process manufacturing executives to focus more on avoiding false negatives, because they can cause a problem to go unnoticed while it is still relatively minor and easy to fix, potentially resulting in a serious crisis.


However, a false alarm can be equally harmful to productivity and profitability in a process plant, and can result in:

  • Alert fatigue, when a team learns to ignore an alert, rather than investigating it, because the alert has often been false in the past.
  • Inflated costs, when parts are replaced unnecessarily and/or production is stopped for unnecessary root cause analysis and repairs.
  • Burnout, when incorrect alarms take up so much time and energy that employees aren’t able to address true positives when they arise.
  • Serious safety issues if/when employees stop responding to safety alerts because so many are false.

How can process plants avoid false alerts?

It can be challenging to avoid false alarms. Many process plant employees respond by turning off alerts or lowering the sensitivity for a system that triggers an alert. But this does not improve the false alarm rate; it only reduces the number that cause an alert.


A better approach uses artificial intelligence (AI) and machine learning (ML) to improve the system’s ability to distinguish a real issue from a false one. For example, Precognize uses ML together with human domain knowledge to create a domain map, identify the most important groups of sensors, and aggregate anomalies to produce a more accurate definition of “normal” that reduces the number of false alerts.


For an ML verification system to be effective, it needs to be trained on accurate, trustworthy, high-quality data, otherwise it too will make mistakes far too often. Otherwise, ML verification will only compound the problem. The first step is thus to set up a reliable process for gathering and sharing plant data.

How do process plants benefit from reducing false alarms?

When there is a lower false positive rate, process plant employees can spend more time resolving serious, complex plant issues that drag down productivity. Lowering the number of false alerts also cuts plant downtime, reduces expenses, and increases safety, resulting in happier and more engaged employees, more profitability, and a stronger bottom line.