Knowledge Graph Scheduling for Underused Machine Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional factory scheduling methods rely heavily on operator experience, making it difficult to identify optimal production combinations, particularly with non-frequently-used machines, which can hinder production efficiency.
Innovation Solution
A scheduling method using a knowledge graph that converts initial scheduling solutions into triples, embeds these into a vector space, and applies a scheduling algorithm to generate a target solution, leveraging deep reinforcement learning to optimize machine and product combinations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional manual scheduling based on operator experience is used, then the scheduling process is simple to implement, but it is difficult to identify optimal production combinations and frequently ignores non-frequently-used machines, reducing production efficiency
Solution Approach 1:
The patent introduces a knowledge graph as an intermediary between the initial scheduling solution and the optimization process. The knowledge graph captures relationships between machines, materials, and products, enabling the system to discover correlations that manual scheduling cannot identify. This intermediary structure allows non-frequently-used machines to be considered in scheduling decisions while maintaining a manageable system architecture.
Solution Approach 2:
The patent replaces manual operator experience with an automated knowledge graph embedding system. Instead of relying on human operators to identify optimal production combinations, the system uses knowledge graph techniques to automatically model and analyze relationships between production resources, thereby discovering optimal schedules including those involving non-frequently-used machines.
2Productivity
If knowledge graph embedding techniques are used to optimize scheduling, then optimal production combinations can be identified including non-frequently-used machines, but the computational complexity and processing requirements increase
Solution Approach 1:
The patent performs preliminary action by constructing the knowledge graph and generating embedded vectors before the actual scheduling optimization. This pre-processing step organizes all relevant relationships between machines, materials, and products in advance, allowing the scheduling algorithm to work with pre-computed representations rather than raw data, thereby reducing real-time computational complexity.
Solution Approach 2:
The patent transforms the scheduling problem from discrete machine selection to continuous vector space optimization. By embedding production resources into vector representations and computing distances in this continuous space, the system can efficiently identify optimal combinations including non-frequently-used machines without exhaustively evaluating all possible discrete configurations.
3Measurement precision
If reinforcement learning or knowledge graph methods are applied, then scheduling accuracy and production yield improve, but the implementation time and computational resources required increase
Solution Approach 1:
The patent performs preliminary action by pre-computing knowledge graph embeddings and storing them for rapid retrieval during scheduling. This allows the system to achieve high scheduling accuracy through sophisticated analysis while reducing implementation time by avoiding repeated computation of the same relationships during the actual scheduling process.
Data Source
AI summary
A scheduling method using a knowledge graph (KG) performs following steps by a processor: obtaining an initial scheduling solution of the production system; converting the initial scheduling solution into triples to form the KG, each triples includes two entities and a relationship of the production system, the two entities indicate two production resources; embedding triples into a vector space to generate embedded vectors by a KG embedding technique; generating embedded vector combinations according to the embedded vectors and computing a distance of each of embedded vector combinations; and performing a scheduling algorithm to generate a target scheduling solution according to the embedded vector combinations, and providing reference information when the scheduling algorithm generates a schedule of a station, wherein the reference information comprises at least one of the embedded vector combinations associated with the station, and the distance corresponding to said at least one embedded vector combinations.