time-delayed partial cross mapping (TDPCM)

Gold definitionUpdated Apr 2, 2026

Definition

Time-delayed partial cross mapping (TDPCM) is a causal inference method designed for direct causality in complex industrial processes. It's part of a causal feature selection framework, utilizing state space reconstruction to analyze interdependent variables and account for time-delayed causal relationships.

At a glance

Executive summary

Time-delayed partial cross mapping (TDPCM) is a method to find direct cause-and-effect links in industrial systems, especially when variables influence each other over time. It improves the reliability of "soft sensors" by accurately identifying these complex relationships, which traditional methods often miss.

TL;DR

TDPCM helps figure out direct cause-and-effect relationships in complicated industrial systems, even when things happen with a delay and are all connected.

Key points

Infers direct causal relationships by employing state space reconstruction to handle interdependent variables and considering varying causal strength across time delays.
Overcomes limitations of traditional causal feature selection methods that ignore time delays and interdependencies, leading to more accurate and stable soft sensor models.
Used by researchers and engineers in process monitoring, soft sensor modeling, and industrial applications requiring robust causal feature selection.
Distinguishes from Time-Delayed Convergent Cross Mapping (TDCCM) by focusing on *direct* causal inference, as opposed to *total* causal inference.
Advances causal inference techniques for complex, dynamic industrial processes to improve model performance and reliability.

Use cases

Developing more accurate and stable virtual sensors for unmeasurable process variables in chemical plants by identifying direct causal inputs with appropriate time delays.
Pinpointing the direct causes of anomalies or deviations in manufacturing processes, enabling faster and more precise fault diagnosis and intervention.
Identifying direct causal factors with time delays that lead to equipment failure in industrial machinery, allowing for more effective scheduling of predictive maintenance activities.
Understanding the direct causal impact of control variables on key performance indicators in complex industrial systems to fine-tune operational parameters for optimization.

Also known as

TDPCM