Federated AI Asset Performance Management for Secure Data Collaboration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional PLM systems face challenges in securely sharing sensitive data across multiple parties, particularly in circular economies, leading to inefficient data collaboration and processing, which hinders effective product lifecycle management and repurposing.
Innovation Solution
A method and system utilizing federated learning and AI models to enable secure data collaboration and decentralized data processing, allowing organizations to train models on local data without sharing sensitive information, ensuring data privacy and improving model performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If sensitive data is shared across multiple parties for collaborative product lifecycle management, then data collaboration and model performance improve, but data security and privacy protection deteriorate
Solution Approach 1:
The patent segments the centralized data processing system into distributed edge devices that process data locally. Each edge device maintains its data locally while collaborating with other edge devices through federated learning, thus segmenting the system to preserve data security while enabling collaboration.
Solution Approach 2:
The patent introduces federated learning as an intermediary mechanism that enables collaborative model training without direct data sharing. The federated learning server coordinates the training process by aggregating model updates from multiple edge devices, acting as a mediator that facilitates collaboration while preserving data privacy.
2Measurement precision
If large volumes of in-use data are collected and processed for performance optimization, then predictive accuracy improves, but infrastructure complexity and processing costs increase
Solution Approach 1:
The patent extracts the data processing function from centralized infrastructure and relocates it to distributed edge devices. By taking out the processing capability from the central server and placing it at the edge, the system reduces infrastructure complexity while maintaining predictive accuracy through local data utilization.
Solution Approach 2:
The patent enables edge devices to perform self-service data processing and model training using their local computational resources. Each edge device independently processes its own in-use data and contributes to the federated learning process, eliminating the need for complex centralized processing infrastructure.
3Reliability
If complex encryption mechanisms are implemented to protect data privacy, then data security improves, but data effectiveness and model training performance deteriorate
Solution Approach 1:
The patent inverts the traditional approach by not encrypting the data itself but rather encrypting the model updates during federated learning. This inversion allows the raw data to remain unencrypted and effective for local processing while still providing security protection through encrypted model parameter transmission.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
The present invention relates a method and system for managing performance of assets (110A-N) in computing environment (100). The method includes receiving a request for managing performance of assets (110A-N) from user devices (108AN). Further, the method includes determining target devices (106A-N) responsible for managing the performance of the assets (110A-N). Furthermore, the method includes determining a baseline Artificial Intelligence (AI) model based on model parameters, asset parameters, and a target device parameter. Further, the method includes classifying the baseline AI model into sub models. The method includes deploying the sub models at each of the target devices (106A-N). Also, the method includes obtaining a real-time data corresponding to the assets (110A-N) from the target devices (106A-N). Additionally, the method includes predicting performance parameters associated with the assets (110A-N) based on the real-time data. Further, the method includes determining actions to be performed on the assets (110A-N).