Production Network Models With Directed-Arc Search Heuristics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing supply chain optimization methods, particularly Mixed Integer Programming (MIP) models, fail to encode the directionality of supply and demand, leading to inefficiencies and limitations in solving complex supply chain problems due to constraints on the number of variables and computational time, making it difficult to achieve both generality and speed in optimization.
Innovation Solution
A network solver approach that translates MIP models into a production network model using production nodes and directed arcs, enabling techniques like backwards-planning and forwards-planning, with a parameterized heuristic and meta-heuristics to improve solution quality and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If MIP modelling techniques are used for supply chain optimization, then a general framework can be established, but the directionality of supply and demand cannot be encoded and computational time becomes excessive
Solution Approach 1:
The patent replaces the MIP optimization framework with a network flow-based system that explicitly encodes directionality through directed arcs and nodes. This substitution enables the system to handle supply chain problems with multiple constraints per variable while maintaining computational efficiency, as network algorithms can process directional flow information natively without requiring complex integer programming formulations.
Solution Approach 2:
The patent segments the supply chain network into distinct nodes (supply nodes, demand nodes, production nodes) and directed arcs representing flow between them. This segmentation allows the system to handle complex problems with multiple constraints by breaking them down into manageable network components, each can be processed efficiently using network algorithms rather than requiring global MIP optimization.
2Manufacturing precision
If standard MIP algorithms are used to solve large supply chain models, then a comprehensive optimization can be achieved, but the models are too large to be solved in a useful amount of time
Solution Approach 1:
The patent substitutes MIP algorithms with network flow algorithms that are specifically designed to handle directional flow problems. This replacement enables the system to solve large-scale supply chain models much faster while maintaining solution quality, as network algorithms can exploit the inherent directional structure of supply chain data to achieve efficient computations without the computational burden of general MIP formulations.
3Speed
If efficient network algorithms are used for supply chain planning, then computational speed is improved, but they cannot run on standard MIP models due to lack of directionality encoding
Solution Approach 1:
The patent creates a universal network model framework that can handle various supply chain planning problems with different constraint structures. By encoding directionality through directed arcs and producing nodes, the system achieves multi-functionality, allowing network algorithms to process a broad range of supply chain problems that would otherwise require problem-specific MIP formulations, thus achieving both speed and generality.
4Productivity
If existing techniques translate MIP models to network models, then some problems can be solved, but they cannot handle variables with more than two constraints
Solution Approach 1:
The patent introduces production nodes that segment the constraint handling process. Each production node can accommodate multiple constraints by representing them as separate arcs entering and leaving the node. This segmentation allows the network model to handle variables with more than two constraints by distributing constraint representations across multiple arcs and nodes, thereby extending the system's versatility without sacrificing productivity.
Data Source
AI summary
Disclosed herein are methods and systems for solving a production network model by iteratively cascading supply decisions. Upstream supply decisions are fixed iteratively, and cascaded down the network in a general way that is independent of any particular use case.