Contextual Hybrid Digital Twin for Accurate Process Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional manufacturing systems face challenges in real-time data processing and creating accurate models due to variations in production systems, making it difficult to monitor and optimize processes effectively, especially with factors like machine state, operator training, and environmental conditions affecting the accuracy of generic models.
Innovation Solution
A processor-implemented method and system that obtain data from multiple sources, train models using physics-based and machine learning models, determine gaps between simulated and real-time data, learn behavioral patterns, and create a contextual hybrid digital twin for real-time monitoring and control of manufacturing processes using soft sensing devices and performance analytics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If generic models are used for manufacturing process monitoring, then model development is simplified, but model accuracy deteriorates due to variations in machine state, operator training, and environmental factors
Solution Approach 1:
The patent segments the manufacturing system into multiple digital twins, each representing a specific machine, operator, and environmental context combination. This segmentation allows models to be tailored to specific conditions rather than using a single generic model, thereby improving accuracy while maintaining manageable complexity through modular structure.
Solution Approach 2:
The patent implements dynamic digital twins that continuously learn and adapt to changing machine states, operator training levels, and environmental factors. This dynamic adaptation enables the system to maintain high accuracy across varying conditions without requiring complete model redesign, balancing simplicity and precision.
2Measurement precision
If real-time data from multiple sources is processed, then monitoring accuracy is improved, but data processing complexity and time consumption increase
Solution Approach 1:
The patent performs preliminary data processing and feature extraction during data collection phases, preparing data in advance for faster real-time analysis. This preliminary action reduces the computational burden during critical real-time monitoring, maintaining accuracy while reducing processing time.
Solution Approach 2:
The patent creates digital twin copies of the physical manufacturing system that replicate complex computations. Once a digital twin is trained, it can rapidly process real-time data without repeating the full training computation, significantly reducing real-time processing time while maintaining monitoring accuracy.
3Productivity
If digital twin environment is connected to control system, then process optimization capability is improved, but system complexity increases
Solution Approach 1:
The patent introduces a control interface layer that acts as an intermediary between the digital twin environment and the physical control system. This intermediary manages the connection, translating between digital simulations and physical control commands, thereby enabling process optimization while managing system complexity through a structured interface architecture.
4Loss of information
If physics-based models are used, then model interpretability is improved, but ability to capture complex real-world behaviors deteriorates
Solution Approach 1:
The patent creates composite digital twin models that combine physics-based models with data-driven machine learning models. The physics-based components provide interpretability and fundamental understanding, while the data-driven components capture complex real-world behaviors that physics models alone cannot represent, achieving both interpretability and accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This disclosure relates generally to systems and methods for monitoring and controlling a manufacturing process using contextual hybrid digital twin. Data pertaining to the manufacturing process is obtained from a plurality of data generation sources which is further inputted to one or more physics based models and train machine learning models such that simulated data and real time tata is obtained. Further, a gap between the simulated data and the real time data is determined and learnt. The learnt gap is further minimized and an augmented set of models are obtained. The augmented set of models along with a set of soft-sensing data is used to create the contextual hybrid digital twin for the manufacturing process. The performance of the manufacturing process is monitored and controlled using a performance analytics and decision making enablers of the contextual hybrid digital twin respectively in real time.