Abstract:We provide some basic and sensible definitions of different types of digital twins and recommendations on when and how to use them. Following up on our recent publication of the Learning Causal Digital Twin, this article reports on a stochastic formulation and solution of the problem. Structural Vector Autoregressive Model (SVAR) for Causal estimation is recast as a state-space model. Kalman filter (and smoother) is then employed to estimate causal factors in a system of connected machine bearings. The previous neural network algorithm and Kalman Smoother produced very similar results; however, Kalman Filter/Smoother may show better performance for noisy data from industrial IoT sources.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: How to use Causal Digital Twin (CDT) technology to understand and quantify the causal relationships between assets in the Industrial Internet of Things (IoT) environment, so as to achieve more effective predictive analysis and decision - making support. Specifically, the paper focuses on the following aspects: 1. **Define and classify different types of digital twins**: - **Display Digital Twin**: Provide real - time visibility of the current state. - **Forward Digital Twin**: Based on physical model simulation, mainly used in the product design stage. - **Causal Digital Twin (CDT)**: Capture the dynamic causal relationships between assets, used to understand system behavior and its potential impacts. 2. **Introduce the stochastic formulation method of causal estimation**: - Reformulate it as a state - space model using the Structural Vector Autoregressive Model (SVAR). - Estimate the causal factors in the system through the Kalman Filter and its variant, the Kalman Smoother. 3. **Handle noisy data**: - The paper specifically points out that for the noisy data common in industrial IoT applications, the Kalman Filter/Smoother may perform better compared to previous neural network algorithms. 4. **Practical application cases**: - Conduct an empirical study using the NASA bearing data set to show how to analyze and predict machine bearing failures through Causal Digital Twin technology. 5. **Improve business results**: - Emphasize the potential of Causal Digital Twin in enhancing production efficiency, improving product quality, and reducing waste, and ultimately increasing the enterprise's gross profit margin. In summary, this paper aims to improve the understanding and prediction ability of complex industrial systems by introducing new causal estimation algorithms (such as the Kalman Filter), especially in cases where causal relationships are clear, and provide strong support for decision - making in the industrial Internet of Things environment.

Stochastic Formulation of Causal Digital Twin: Kalman Filter Algorithm

Digital Twins in Engineering Dynamics: Variational Equation Identification, Feedback Control Design and Their Rapid Update

Structural & Granger CAUSALITY for IoT Digital Twin

Model-Based Monitoring and State Estimation for Digital Twins: The Kalman Filter

Machine learning based digital twin for stochastic nonlinear multi-degree of freedom dynamical system

Causal Falsification of Digital Twins

Design of Digital Twin Sensing Strategies Via Predictive Modeling and Interpretable Machine Learning

Causal Semantic Communication for Digital Twins: A Generalizable Imitation Learning Approach

Deep Kalman Filters Can Filter

A Probabilistic Graphical Model Foundation for Enabling Predictive Digital Twins at Scale

Machine learning based digital twin for dynamical systems with multiple time-scales

Smoothing Dynamic Systems with State-Dependent Covariance Matrices

Causal Time-Series Synchronization for Multi-Dimensional Forecasting

The role of surrogate models in the development of digital twins of dynamic systems

A Collaborative Kalman Filter for Time-Evolving Dyadic Processes

An Application of a Digital Twin to Robotic System Design for an Unstructured Environment

Digital Twin Data Modelling by Randomized Orthogonal Decomposition and Deep Learning

Digital Twin-Based Multiple Access Optimization and Monitoring via Model-Driven Bayesian Learning

Kalman Filtering of Distributed Time Series

Digital twins of nonlinear dynamical systems: A perspective