Flexible Job Shop Scheduling With Hypergraph Policy Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing flexible job shop scheduling methods face challenges in handling complex and variable scheduling conditions, particularly in large-scale systems, due to difficulties in modeling and scalability, leading to inefficiencies and suboptimal solutions.
Innovation Solution
A hypergraph-based method using a hypergraph neural network to extract job features and combine them with policy networks to determine actions that change the scheduling state, incorporating k-nearest neighbors graphs to reduce candidate actions and improve computational efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional job shop scheduling methods are used, then the system is simple to implement, but it cannot effectively handle complex and variable scheduling conditions
Solution Approach 1:
The scheduling problem is segmented into discrete job tasks, operation steps, and machine operations. Each job task is broken down into sequential operation steps, and each operation step is assigned to specific machines. This segmentation allows the system to handle complex scheduling conditions by processing individual components rather than the entire system at once, improving adaptability while maintaining manageable complexity.
Solution Approach 2:
The scheduling system implements dynamic decision-making through the determination of actions that change the shop scheduling state. The system can adapt to varying scheduling conditions by dynamically selecting which job task to process next and which machine to assign, allowing flexibility in response to changing priorities, machine availability, and job requirements without requiring a complete redesign of the scheduling framework.
2Adaptability or versatility
If more candidate machines are provided for each job step, then production flexibility improves, but the difficulty of determining optimal machine selection increases
Solution Approach 1:
The system changes the parameters used for machine selection by considering multiple factors simultaneously: current shop scheduling state, job task priorities, machine availability, processing times, and sequential order dependencies. By evaluating machines based on multiple dynamic parameters rather than a single criterion, the system can effectively select from multiple candidate machines while managing the complexity of optimization through comprehensive parameter consideration.
3Manufacturing precision
If comprehensive scheduling constraints are enforced, then solution quality improves, but computation time increases
Solution Approach 1:
The system applies partial action by determining actions that change the shop scheduling state incrementally rather than optimizing the entire schedule at once. Each decision focuses on selecting the next job task and machine assignment based on current state and constraints, achieving good solution quality through a series of locally optimal decisions rather than requiring computationally intensive global optimization.
Solution Approach 2:
The system performs preliminary action by pre-establishing the sequential order dependency relationships between operation steps and pre-identifying candidate machines for each job step. These preliminary structures are prepared in advance, allowing the scheduling system to quickly make decisions by referencing pre-defined constraints and relationships rather than calculating everything from scratch during execution.
Data Source
Figure 1
Figure 2(a)~2(c)
Figure 3
AI summary
A flexible job shop scheduling method is provided. The method includes: obtaining a shop scheduling state including at least one of a sequential order dependency relationship between a plurality of job tasks being processed, a sequential order dependency relationship between a plurality of operation steps in each job task of the plurality of job tasks, a processing/being processed relationship between the plurality of job tasks and a plurality of machines, and a mutual constraint relationship between the plurality of machines; representing the shop scheduling state as a state hypergraph using a hypergraph; extracting a hypergraph-based job feature from the state hypergraph using a hypergraph neural network; and determining an action to change the shop scheduling state according to the hypergraph-based job feature by using a policy network.