Distributed Learning Servers for Manufacturing Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current manufacturing data processing systems for FA devices like machine tools and robots are expensive due to the need for ultrafast networks, large databases, and high-performance processors to handle real-time data, and suffer from poor cost efficiency, especially when not all devices are operational daily and data generation varies.
Innovation Solution
A manufacturing data processing system with a management device that dynamically determines the combination of data processing devices and communication channels based on speed and capability, including a network management device for switching connections, to optimize data processing and reduce costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a centralized learning server with large database and high-performance processor is used to handle real-time data from multiple FA devices, then data processing capability is improved, but hardware costs and system complexity increase significantly
Solution Approach 1:
The patent segments the centralized learning server into multiple distributed learning servers, each handling a subset of FA devices. This distribution reduces the processing burden on each individual server, lowering hardware requirements and system complexity while maintaining overall data processing capability across the networked system.
Solution Approach 2:
The patent introduces a new dimension of spatial distribution by deploying learning servers across multiple locations in the network rather than concentrating all processing power in a single centralized server. This dimensional change from centralized to distributed architecture reduces the complexity and hardware requirements of each node while preserving total system capability.
2Adaptability or versatility
If hardware resources are provided in each FA device to perform local learning, then learning can be performed autonomously, but hardware costs increase
Solution Approach 1:
Instead of providing full learning capabilities in each FA device, the patent implements partial action by enabling selected FA devices to perform specific learning tasks locally while relying on distributed learning servers for more complex processing. This selective approach maintains autonomous learning capability where needed while avoiding the excessive hardware costs of equipping every device with complete learning resources.
Solution Approach 2:
The patent creates a multi-functional system where learning servers serve multiple FA devices simultaneously, and selected FA devices can both generate data and perform local learning. This universality allows the same infrastructure to support both centralized and distributed learning modes, reducing overall hardware requirements while maintaining autonomous capability.
3Device complexity
If a learning server is provided and connected through network to FA devices, then hardware costs are reduced, but communication overhead and network dependency increase
Solution Approach 1:
The patent applies local quality by enabling selected FA devices to perform learning locally without network communication, while other devices continue to use the distributed learning servers. This creates different communication qualities for different devices based on their specific needs, reducing overall network dependency and improving reliability for critical operations.
Solution Approach 2:
The patent introduces communication protocols and data formats as intermediaries that standardize interactions between FA devices and learning servers, reducing communication overhead. These intermediaries enable efficient data exchange while minimizing network dependency, allowing the system to maintain reduced hardware costs while improving communication reliability.
4Speed
If ultrafast network and extremely large database are used to process real-time data from hundreds or thousands of FA devices, then data processing speed is improved, but system cost increases very much
Solution Approach 1:
The patent segments the data processing load across multiple distributed learning servers, each handling a portion of the real-time data stream. This segmentation allows the system to process data from hundreds or thousands of FA devices without requiring a single ultrafast network connection or extremely large database, thereby reducing system cost while maintaining processing speed through parallel operation of multiple servers.
Solution Approach 2:
The patent implements dynamic load balancing and data routing that adapts to varying data generation rates from different FA devices. This dynamic approach optimizes data processing speed by directing data streams to appropriate learning servers based on current system conditions, maintaining high performance without requiring permanently over-provisioned expensive infrastructure.
Data Source
AI summary
A manufacturing data processing system includes a plurality of manufacturing apparatuses, a plurality of data processing devices for processing manufacturing data associated with the plurality of manufacturing apparatuses, a plurality of communication channels for communicating the manufacturing data between the plurality of manufacturing apparatuses and the plurality of data processing devices, and a management device. The management device determines a combination of the data processing device that processes the manufacturing data associated with each of the plurality of manufacturing apparatuses and the communication channel that communicates the associated manufacturing data between each of the plurality of manufacturing apparatuses and the data processing device, based on the communication speed of the communication channel and the data processing capability of each of the plurality of data processing devices.