Intelligence Graphs for Automatic Extraction Across Product Data Silos
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing product development and manufacturing processes face challenges due to dispersed information across multiple data silos, lacking direct connections and compatibility, leading to inefficiencies and laborious modeling of product variants.
Innovation Solution
A computer-implemented method integrates data from different data silos into an intelligence graph, using nodes and edges to represent relationships, enabling AI-based functions like graph neural networks for comprehensive analysis and prediction of product variants.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data from multiple data silos are kept separate and independent, then each data silo maintains its own data structure and compatibility, but the information required for holistic product perspective is scattered and not directly connected
Solution Approach 1:
The patent merges data from multiple independent data silos into a unified intelligence graph structure. The data integration function combines data from different sources (product development, manufacturing, supply chain) into a single graph where nodes represent data entities and edges represent relationships, enabling holistic product perspective while maintaining data integrity from source systems.
Solution Approach 2:
The intelligence graph acts as an intermediary layer between distributed data silos. The data integration function serves as a mediator that transforms and harmonizes data from different sources into a common graph representation, allowing information flow between previously disconnected systems without direct integration between each silo.
2Measurement precision
If specialized tools are used for product development and optimization, then domain professionals gain in-depth insights into specific characteristics, but the information is scattered across several data silos with no direct connection between them
Solution Approach 1:
The intelligence graph provides a universal data structure that can represent multiple dimensions of product information simultaneously. A single graph can accommodate data from various specialized tools (development, manufacturing, supply chain) using standardized node and edge types, enabling multi-functional access to diverse data sources without requiring separate specialized systems for each domain.
3Loss of information
If data from distributed data silos are integrated into a unified structure, then a holistic product perspective is achieved, but compatibility between data formats and structures across different data silos must be ensured
Solution Approach 1:
The data integration function transforms data from various sources by changing its parameters into a standardized graph format. Different data formats and structures from source systems are converted into uniform node and edge representations in the intelligence graph, with standardized attributes and relationship types that maintain compatibility across diverse input sources.
4Productivity
If optimization is performed in one expert tool, then specific product characteristics are improved, but it is not synchronized with another tool, potentially leading to discrepancies or misinterpretations
Solution Approach 1:
The intelligence graph provides a centralized feedback mechanism that synchronizes optimization changes across all connected data silos. When product characteristics are optimized in one tool, the changes are reflected in the graph structure and automatically propagated to other connected systems, ensuring data consistency and preventing discrepancies through real-time or near-real-time synchronization.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
System and computer-implemented method for automatically extracting and evaluating knowledge from distributed data silos (90, 91, 92) regarding a product manufactured in a production plant, comprising: - receiving (S1), by a data interface (81), data of at least two different data silos (90, 91, 92) comprising data of different dimensions with respect to the product, wherein at least one data silo (90, 91, 92) comprises a bill of materials of the product, - assigning (S2), by a data integration function (82), the data of all data silos (90, 91, 92) to an intelligence graph (10), comprising nodes and edges connecting two nodes, wherein each data point of the data is assigned to one node, and each edge denotes a relationship between the interconnected nodes, - identifying (S3), by a graph processing function (83) receiving as input the intelligence graph (10), a subgraph (30, 40, 50, 60) depending on a task related to the product, wherein the subgraph (30, 40, 50, 60) comprises information according to the task, and - outputting (S4), by a user interface (84), the information and/or the subgraph (30, 40, 50, 60) according to the task.