融合LLM和CNL的船舶元件模板自动生成方法  

Automatic Generation Method of Ship Element Template by Integrating LLM and CNL

在线阅读下载全文

作  者:曹金浩 宋元斌[1] CAO Jin-hao;SONG Yuan-bin(School of Ocean and Civil Engineering,Shanghai Jiao Tong University,Shanghai 200240)

机构地区:[1]上海交通大学船舶海洋与建筑工程学院,上海200240

出  处:《制造业自动化》2025年第2期132-139,共8页Manufacturing Automation

摘  要:传统的船舶元件模板的编制一般由设计人员提出需求,由软件开发人员编写代码。由于软件开发人员缺少船舶设计的专业知识,而设计人员缺少计算机编程知识,导致双方沟通难度大。为解决上述问题,提出了一种融合大语言模型(LLM)和受控自然语言(CNL)的船舶元件模板自动生成方法。通过LLM将自然语言(NL)描述的建模要求转为CNL建模要求,再根据专门设计的映射规则将CNL建模要求转换为元件模板的可视化编码。200个船舶管系典型元件模板的实验结果表明,所提出的模板自动生成方法具有较高准确率,可以用于实船设计工作,同时上述方法可以避免高成本的LLM微调。另外,上述方法全过程不需要软件开发人员参与,解决了设计人员与编程人员的沟通困难,建模效率显著提升。The modeling of ship element template is traditionally collaborated between the ship designers who put forward requirements and the software developers who write codes.The lack of expertise in ship design among software developers and the lack of computer programming knowledge among designers,however,tend to make the communication between the two parties difficult.To address this issue,an automatic generation method of ship element template is developed by integrating the large language model(LLM)and the controlled natural language(CNL).The instructions for modeling an element template can be easily described in natural language(NL),and then converted into CNL clauses using LLM.Thereafter,the CNL requirements are converted into visual coding utilizing specially designed mapping rules.The experimental results of 200 typical templates of ship piping elements show that the proposed method has a high accuracy and can be used in real ship design.At the same time,the developed method can avoid the costly LLM fine-tuning.In addition,the entire process of the above method does not require the participation of programmers,which solves the communication difficulties between designers and programmers,and therefore can significantlt improve the modeling efficiency.

关 键 词:元件模板 大语言模型 受控自然语言 嵌入式向量 语义相似度 可视化编码 

分 类 号:U674.11[交通运输工程—船舶及航道工程]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象