Getting the jump on industrial failures

Modern automated industrial processes rely heavily on the precise control of process conditions, making it critical to detect emerging deviations. A KAUST-led research team developed a highly sensitive incipient anomaly detection method with the potential to dramatically improve industrial productivity, quality and safety.

From oil and gas refining to water treatment and product manufacturing, precise process control has become an integral but largely hidden part of modern industry. However, keeping an automated process running smoothly and safely and producing the desired results remains a major challenge in many sectors. Even small deviations in process parameters can result in lost time, and catastrophic failure can bring devastating health, safety and financial consequences. Because of this, engineers must keep tweaking and improving the reliability of their processes, watching carefully for signs of anomalies that could lead to disaster.

“The detection of incipient anomalies is crucial to maintain the normal operation of a system by providing an early warning,” noted Ying Sun, assistant professor of applied mathematics and computational science from the University’s Environmental Statistics group. “The problem is that incipient anomalies are often too weak to be detected by conventional monitoring methods.”

Conventional methods for detecting process variations rely on the statistical identification of trends in the constant stream of monitoring data from key points in the process. To avoid false alerts, however, many observations are needed before a warning is triggered, so when an anomaly is detected, it is often too late to avert a major problem, demanding a process shutdown.

“Our method incorporates information from the entire process history, rather than just the most recent observations, so that the detection indicator is more sensitive to small changes,” explained Sun.

The team’s statistics-based approach involves construction of memory chart of multiple monitoring parameters using a statistical technique known as principle component analysis. The test statistics for new measurements are then compared with a control limit set using healthy historical process data for anomaly detection and decision-making.

“Another advantage of our scheme is that it can be easily implemented in real time because of its low computational cost,” Sun said.

Using their incipient anomaly detection scheme, the team detected very weak emerging anomalies in an air flow heating system that could not be detected by conventional methods.

“Statistical quality control methods have a wide range of applications,” added Sun. “At KAUST, we are also developing incipient anomaly detection methods to monitor wastewater treatment plants and water desalination plants.”

References

1. Harrou, F., Madakyaru, M., Sun, Y., & Khadraoui, S. Improved detection of incipient anomalies via multivariate memory monitoring charts: Application to an air flow heating system. Applied Thermal Engineering 109, 65-74 (2016).| article