Decentralized Fleet Analytics Using Local ML Model Sharing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Centralized fleet analytics approaches incur high network traffic and storage costs, and sensitive data sharing issues among distinct entities, particularly when plants are owned by individual customers.
Innovation Solution
A decentralized approach where each plant builds a machine learning model from local data streams, transmitting the model to a cloud server for ensemble learning, thereby reducing data transmission and allowing entities to benefit from collective knowledge without sharing sensitive data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all data from individual plants is sent to a centralized cloud server for analysis, then comprehensive fleet-wide machine learning models can be derived, but network traffic and storage costs increase significantly
Solution Approach 1:
The system segments the centralized analytics approach by allowing each plant to build its own local machine learning model using its data, rather than sending all data to a central server. This segmentation reduces network traffic while maintaining predictive capabilities through decentralized model training and periodic cloud-based model updates.
Solution Approach 2:
The system extracts only the essential elements (machine learning models) from the data processing workflow and transfers them between cloud and edge, rather than moving all raw data. This extraction approach maintains analytical benefits while minimizing data transmission requirements.
2Measurement precision
If all data from individual plants is sent to a centralized cloud server for analysis, then comprehensive fleet-wide machine learning models can be derived, but storage costs increase significantly
Solution Approach 1:
The system segments the data processing workflow so that each plant maintains its own data locally and builds its own model, avoiding the need to store all plant data centrally. Only model parameters and aggregated statistics are stored in the cloud, dramatically reducing storage requirements.
Solution Approach 2:
Instead of storing and transmitting all raw data, the system creates and transmits copies in the form of machine learning models. These models capture the essential patterns and insights from the data without requiring the actual data to be stored or transmitted, reducing storage and network costs.
3Adaptability or versatility
If data is sent to a centralized cloud server, then fleet-wide analytics can be performed, but entities may be unwilling to share sensitive data
Solution Approach 1:
The system segments the analytics process so that sensitive raw data remains localized at each plant and is never transmitted to the cloud. Each entity maintains full control of its own data while still benefiting from fleet-wide analytics through the exchange of anonymized model parameters and aggregated statistics.
Solution Approach 2:
The machine learning model acts as an intermediary that transforms sensitive plant-specific data into useful fleet-wide insights without exposing the underlying sensitive information. The model captures patterns and predictions while preserving data privacy, enabling collaboration without direct data sharing.
Data Source
AI summary
Technologies for decentralized fleet analytics are disclosed. In at least one embodiment, a local cloud service at a plant site builds a first machine learning model of one or more first streams of data associated with the plant site. The local cloud service sends the first machine learning model to a cloud service connected to the plant site and other plant sites. The local cloud service receives a second machine learning model from the cloud service. The second machine learning model is trained as a function of the first machine learning model and one or more machine learning models built by the other plant sites. The local cloud service updates the first machine learning model based on the second machine learning model.