time-delayed convergent cross mapping (TDCCM)

Gold definitionUpdated Apr 2, 2026

Definition

Time-delayed convergent cross mapping (TDCCM) is a causal feature selection method designed for industrial processes. It addresses time-delayed and interdependent variable relationships by employing state space reconstruction to identify total causal inference, enhancing soft sensor model accuracy and stability.

At a glance

Executive summary

Time-delayed convergent cross mapping (TDCCM) is a method to find cause-and-effect relationships in industrial systems more accurately. It's special because it accounts for how long it takes for a cause to have an effect and how different parts of a system influence each other, which helps make better prediction models.

TL;DR

TDCCM is a technique that figures out cause-and-effect in complicated industrial systems by considering time delays and how everything is connected, making predictions more reliable.

Key points

Employs state space reconstruction to analyze causality, considering varying causal strength across time delays.
Solves the problem of traditional causal inference methods ignoring time delays and interdependence in industrial processes, improving soft sensor model accuracy and stability.
Used by researchers and engineers in process monitoring, industrial automation, chemical engineering, and data-driven modeling.
Unlike traditional causal inference methods that assume simultaneous causality and decorrelation, TDCCM explicitly models time-delayed and interdependent relationships.
Part of a broader trend in developing more robust and realistic causal inference methods for complex, dynamic systems, especially in industrial AI and digital twins.

Use cases

Chemical Process Optimization: Identifying the time-delayed causal factors affecting product quality or yield in a chemical reactor to optimize operating conditions.
Manufacturing Anomaly Detection: Pinpointing the root causes of equipment failures or production defects by analyzing time-delayed causal links between sensor readings and machine states.
Environmental Monitoring: Determining the causal influence of various environmental parameters (e.g., temperature, humidity, pollutant levels) on each other with appropriate time lags.
Energy Management: Optimizing energy consumption in smart factories by understanding the time-delayed causal relationships between production schedules, machine operations, and energy demand.

Also known as

TDCCM, Time-delayed cross mapping, Convergent cross mapping, TDPCM