Wednesday 16 April 2025
The quest for a more effective way to monitor and control complex industrial processes has led researchers to develop a new approach that combines statistical process control with machine learning techniques. By integrating these two methods, scientists have created a framework that can detect anomalies in real-time and provide accurate fault diagnosis.
The new system, called FARM (Fault Detection and Classification using Statistical Process Control and Riemannian Geometry Analysis), has been tested on the Tennessee Eastman process – a benchmark problem used to evaluate the performance of various monitoring methods. The results show that FARM outperforms existing approaches in terms of speed, accuracy, and robustness.
One of the key innovations behind FARM is its use of statistical process control (SPC) to detect anomalies in the data stream. SPC is a widely used method for monitoring industrial processes, but it can be limited by its reliance on traditional statistical methods that are not well-suited to handling large amounts of complex data. FARM addresses this limitation by incorporating machine learning techniques, such as Riemannian geometry analysis, into the SPC framework.
Riemannian geometry analysis is a powerful tool for dealing with high-dimensional data, which is common in industrial processes. By mapping the covariance matrices of the process data to a Riemannian manifold, FARM can identify patterns and relationships that would be difficult or impossible to detect using traditional statistical methods.
The combination of SPC and Riemannian geometry analysis allows FARM to detect anomalies in real-time, even when they are subtle or occur infrequently. This is because the system is able to learn from historical data and adapt to changing conditions over time.
In addition to its ability to detect anomalies, FARM also provides accurate fault diagnosis by using a modified support vector machine (SVM) algorithm. The SVM algorithm is trained on a dataset of known faults and uses this training to identify unknown faults in the process data.
The performance of FARM has been evaluated using a range of metrics, including fault detection rate, false alarm rate, and average classification accuracy. The results show that FARM outperforms existing approaches in all three categories, with an average fault detection rate of 96.97% and an average false alarm rate of 0.0036.
The potential applications of FARM are vast and varied. It could be used to monitor and control a wide range of industrial processes, from chemical plants to power grids.
Cite this article: “Boosting Process Monitoring with Machine Learning: A Novel Approach to Fault Detection and Classification”, The Science Archive, 2025.
Industrial Process Monitoring, Fault Detection, Machine Learning, Statistical Process Control, Riemannian Geometry Analysis, Anomaly Detection, Real-Time Monitoring, Fault Diagnosis, Support Vector Machine, Process Control.
