Contextual Transformation of Analytical Models for IIoT Edge Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial Internet of Things (IIoT) edge nodes often lack the necessary software applications to support dynamically connected devices, and existing analytical models trained in the cloud may not be directly applicable due to resource limitations and varying execution runtimes, making it difficult to adapt models for specific edge environments.
Innovation Solution
The method involves receiving an analytical model from a cloud service, obtaining local data from the IIoT edge node, analyzing this data to determine the situational context, and transforming the model accordingly to create a transformed analytical model that is compatible with the edge environment, enabling execution on diverse hardware platforms without altering the model's functionality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If analytical models are persisted on edge nodes to ensure availability, then model accessibility is improved, but resource consumption increases
Solution Approach 1:
The patent applies local quality by transforming analytical models to have different characteristics suitable for edge node execution. The model transformation process adapts cloud-trained models to edge-specific contexts, creating locally optimized versions that consume fewer resources while maintaining reliability. This is achieved through contextual transformation that adjusts model parameters and structures based on edge node capabilities.
Solution Approach 2:
The patent utilizes parameter changes by modifying model parameters during the transformation process from cloud to edge deployment. The system changes execution parameters, data formats, and model configurations to optimize for edge node resources. This includes adjusting model complexity, data sampling rates, and computation parameters to balance reliability and resource consumption.
2Loss of time
If cloud-trained analytical models are deployed directly to edge nodes, then deployment speed is improved, but model adaptability deteriorates
Solution Approach 1:
The patent applies preliminary action by performing model transformation in advance during the deployment process. Rather than deploying cloud models directly and then adapting them later, the system pre-transforms models to edge-specific contexts before deployment. This preliminary transformation ensures both rapid deployment and immediate adaptability to edge environments.
Solution Approach 2:
The patent uses an intermediary transformation process between cloud model training and edge deployment. The model transformation service acts as a mediator that converts cloud-trained models into edge-compatible formats while preserving adaptability. This intermediary step enables fast deployment by preparing models beforehand while maintaining versatility through contextual adaptation.
3Adaptability or versatility
If edge nodes support multiple device types dynamically, then system versatility is improved, but device complexity increases
Solution Approach 1:
The patent applies universality by creating transformed analytical models that can serve multiple device types on edge nodes. The model transformation process generates universal models that adapt to different device contexts without requiring separate models for each device type. This enables edge nodes to support diverse devices dynamically while maintaining manageable complexity through standardized transformation processes.
Data Source
AI summary
Disclosed are methods, systems, and non-transitory computer-readable medium for a contextual transformation of an analytical model for an industrial internet of things (IIoT) edge node. For instance, the method may include receiving the analytical model from a cloud service; obtaining local data of the IIoT edge node; analyzing the local data to determine a situational context of the IIoT edge node; determining whether to transform the analytical model based on a fit between the analytical model and the situational context; and in response to determining to transform the analytical model, transforming the analytical model based on the situational context to derive a transformed analytical model.