The layout of pipeline supports (LPS) is essential for ensuring the safety of pipeline, which prioritize safety while considering spatial arrangement, pipeline characteristics, and technical constraints. Due to the complexity of the design process, there has been a long-term reliance on manual design, resulting in significant efficiency bottlenecks in the design phase. The research introduces a design framework for LPS based on sequential generation. Initially, a feature vector is defined by focusing on the significant design factors. Subsequently, a knowledge-driven variable-step-size sampling method introduces to optimize efficiency without compromising effectiveness. Then, the sampling sequence is processed through a Multi-Heads-Attention based Encoder (MHA-Encoder) to predict labels for each point in the sequence, which indicating the support type and its index signifying the placement. Finally, the proposed method’s validity was confirmed via real-world design cases, demonstrating an accuracy of 91.55% with minimal errors and a 40-fold reduction in the design cycle.
A sequential-generation-based intelligent method for layout of pipeline supports
2025
Article (Journal)
Electronic Resource
Unknown
Metadata by DOAJ is licensed under CC BY-SA 1.0
A sequential-generation-based intelligent method for layout of pipeline supports
Elsevier | 2025
|Machine Learning Pipeline Supports AI-based Applications
Springer Verlag | 2025
|Model pipeline "supports long-term growth"
Online Contents | 2009