Edge Analytics Architecture for Low-Latency Industrial Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current industrial systems lack the capability to generate predictive and prescriptive actions in real-time, especially in settings with limited network connectivity and noisy sensor data.
Innovation Solution
A distributed analytics system with an edge processing device that performs predictive and prescriptive analytics locally, using an architect subsystem to deploy analytic models and modify them based on system performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If centralized data analysis approach is used, then data processing capability is improved, but latency between sensor data collection and prescription of action increases
Solution Approach 1:
The system segments data processing functions by deploying edge computing devices at multiple levels (edge, fog, cloud) rather than centralizing all processing. Each segment handles appropriate analytics locally, reducing latency while maintaining overall processing capability through distributed architecture.
Solution Approach 2:
The patent introduces a spatial dimension to data processing by distributing analytics capabilities across multiple physical locations and network levels. Instead of a single centralized processing point, the system creates a multi-dimensional processing hierarchy that brings computation closer to data sources, reducing transmission latency.
2Ease of operation
If network connectivity is assumed consistently available, then centralized analysis can be implemented, but system reliability deteriorates in settings with limited connectivity
Solution Approach 1:
The system dynamically adapts its operational mode based on network availability. Edge devices can function autonomously when connectivity is limited, processing data locally and synchronizing with centralized systems when available. This dynamic behavior maintains reliability across varying network conditions while preserving ease of operation through automated adaptation.
Solution Approach 2:
Edge computing devices are equipped with autonomous capabilities to perform data processing and analytics independently of centralized systems. This self-service capability ensures continuous operation during network outages, with devices automatically syncing results when connectivity is restored, thereby maintaining reliability without requiring constant centralized management.
3Measurement precision
If big data models are trained at data center level, then model accuracy is improved, but deployment time and adaptability to real-time operation deteriorate
Solution Approach 1:
The model lifecycle is segmented into distinct phases: heavy training and validation occur at the data center level to ensure accuracy, while lighter model deployment and real-time adaptation occur at edge devices. This segmentation allows high-accuracy models to be developed centrally while maintaining real-time adaptability through distributed deployment and continuous learning at the edge.
Solution Approach 2:
Comprehensive model training, validation, and optimization are performed in advance at the data center level before deployment. This preliminary action ensures high model accuracy is achieved beforehand, allowing edge devices to deploy pre-trained models quickly and adapt them in real-time with minimal computational overhead, thus balancing accuracy with real-time adaptability.
4Speed
If PLC-based real-time control is used, then response speed is improved, but predictive and prescriptive analytics capability deteriorates
Solution Approach 1:
The system merges deterministic PLC control logic with data-driven analytics capabilities into a unified edge computing platform. This integration allows the system to maintain fast real-time control responses from PLC while simultaneously executing predictive and prescriptive analytics workloads, achieving both speed and advanced analytics capability through hardware and software convergence at the edge.
Data Source
AI summary
A distributed analytics system to control an operation of a monitored system, and method of operation thereof, including an architect subsystem and an edge processing device. The edge subsystem includes an edge processing device associated with the monitored system. The architect subsystem is configured to deploy an analytic model to the edge processing device based on characteristics of the monitored system. The edge processing device is configured to receive the analytic model and independently perform predictive and prescriptive analytics on dynamic input data associated with the monitored system, provide control signals to the monitored system according to the predictive and prescriptive analytics, and provide information to the architect subsystem, including monitored system responses to the control signals. The architect subsystem is configured to modify the analytic model to improve system performance of the monitored system.