New rARX-DIPCA Algorithm Tracks Time-Varying Processes in Real Time

A new recursive algorithm, rARX-DIPCA, has been developed by researchers Deepanjhan Das and Shankar Narasimhan for identifying errors-in-variables autoregressive models with exogenous input (EIV-ARX) in time-varying single-input single-output (SISO) processes. The algorithm builds on a recently developed recursive iterative PCA method, enabling real-time adaptation to sensor degradation and changes in model coefficients. It recursively updates model parameters and noise variances as new measurements arrive, without storing historical data beyond a specified lag window. This computational efficiency is achieved through online covariance updates, allowing the system to simultaneously identify process order, time delay, and noise variances.

The work has been accepted for publication at the 23rd IFAC World Congress in Busan, Republic of Korea. Simulation studies on benchmark systems demonstrated effective tracking performance and practical applicability, particularly for industrial control environments where sensors degrade over time. By eliminating the need for batch data storage and reprocessing, rARX-DIPCA enables continuous model identification and adaptation, reducing downtime and improving system reliability. The approach is especially relevant for manufacturing, chemical processes, and power systems where process dynamics evolve over time and sensor accuracy drifts.