Edge Matrix Acceleration for Real-Time Digital Twin Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial automation systems face challenges in deploying computationally expensive techniques like deep neural networks due to inadequate computational resources on factory floors, requiring specialized edge computing capabilities for sensitive and timely industrial automation data processing.
Innovation Solution
Deployment of a hardware accelerator, such as a neural processing unit (NPU), within industrial networks to execute deep neural networks and perform large-scale matrix operations efficiently, enabling rapid execution and real-time updates to digital twin simulation models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If deep neural networks are deployed in industrial automation systems, then intelligent control capabilities are improved, but computational resource requirements increase significantly
Solution Approach 1:
The patent segments the computational workload by deploying deep neural networks on edge devices rather than centralized cloud systems. This distribution allows intelligent control capabilities to be embedded at the factory floor level, enabling real-time decision-making while reducing the burden on centralized computational resources.
Solution Approach 2:
The patent introduces hardware accelerators as intermediary components between standard processors and deep neural network operations. These specialized units (such as NPUs or FPGAs) mediate the computational demands, providing the necessary processing power for intelligent control while being more resource-efficient than general-purpose processors.
2Speed
If computational resources are increased for deep neural networks, then processing speed is improved, but device complexity increases
Solution Approach 1:
Hardware accelerators serve as intermediary components that provide specialized processing power for deep neural networks without requiring a complete redesign of the entire computing system. These accelerators can be integrated into existing edge devices, improving processing speed while minimizing the increase in overall device complexity.
Solution Approach 2:
The patent employs hardware accelerators that can be configured for multiple functions and applications. These multi-functional units can handle various deep neural network operations and can be adapted to different industrial automation tasks, providing high processing speed without requiring separate specialized hardware for each function.
3Loss of time
If edge computing capabilities are deployed on factory floor, then real-time data processing is improved, but infrastructure complexity increases
Solution Approach 1:
The patent extracts the deep neural network computational capabilities from centralized cloud systems and places them directly on edge devices at the factory floor. This extraction enables real-time data processing by eliminating network latency, while the modular nature of edge devices keeps infrastructure complexity manageable.
Solution Approach 2:
Hardware accelerators act as intermediaries that enable real-time processing capabilities on edge devices without requiring complete infrastructure overhaul. These accelerators can be added to existing edge computing platforms, providing the necessary computational power for real-time operations while maintaining relatively simple infrastructure architecture.
Data Source
AI summary
An edge device can be configured to perform industrial control operations within a production environment that defines a physical location. The edge device can include a plurality of neural network layers that define a deep neural network. The edge device be configured to obtain data from one or more sensors at the physical location defined by the production environment. The edge device can be further configured to perform one or more matrix operations on the data using the plurality of neural network layers so as to generate a large scale matrix computation at the physical location defined by the production environment. In some examples, the edge device can send the large scale matrix computation to a digital twin simulation model associated with the production environment, so as to update the digital twin simulation model in real time.