信息融合流模型研究

传统信息融合解决方案存在资源管理概念缺乏、功能固化、技术落后等问题,并导致系统出现功能紧耦合,以及自适应能力差,不能满足信息融合需求的变化.针对存在的问题,以及传统信息融合将业务逻辑与过程逻辑以紧耦合的方式固化于系统中的特点,提出将信息融合中业务逻辑与过程逻辑分离,二者之间建立显式映射规则.以此为基础建立信息融合流模型,探讨了流模型的组成以及相应建模方法以及信息融合系统如何应用信息融合流模型实现信息融合任务.基于流模型的信息融合解决方案将大大增强系统解决能力的提高,增强系统灵活性,从而降低系统开发成本,增强对不断变化的信息融合任务需求的响应敏捷性.     

MMCUP:融合多模态信息的代码注释自动更新方法

良好的代码注释对于程序维护有着重要价值.但在实际开发过程中,开发人员经常会在更改代码后忽略更新相应的代码注释,导致更新后的代码和注释不一致,对软件可维护性造成影响.现有注释更新方法在进行注释更新时,通常仅将代码视为普通文本进行处理,忽视了代码结构信息.为此,本文提出了一种融合多模态信息的代码注释更新方法MMCUP(Multi-Modal Comment UPdating).MMCUP使用了旧代码注释、代码编辑序列和AST差异序列三种模态的信息来训练基于Transformer架构的模型,以对注释进行更新.实验结果表明,MMCUP在Accuracy、Recall@5等指标上相较于CUP和HatCUP等方法至少提高了5.8%和4.4%.     

代码注释自动生成研究进展

代码注释作为软件中的重要组成部分,在软件维护、复用等领域中发挥着重要作用。代码注释自动生成技术旨在减轻人工编写注释的工作量,从而提高软件开发效率。现有的注释自动生成方法分为基于规则、文本摘要、数据驱动、主题模型、深度学习等层次。综述了代码注释自动生成的相关概念,对比总结各类代码注释自动生成方法,对近年来代码注释自动生成相关应用进行了梳理和总结,最后对注释自动生成所面临的挑战进行了分析,展望了该领域未来的研究。     

多专家注释的视杯和视盘不确定性量化

现有基于深度学习的视杯和视盘分割方法在模型训练时,仅使用图像的单个注释或从多个注释中获取唯一的注释信息,忽略原始多专家标注中嵌入的一致性或差异性信息,从而导致模型和预测结果过度自信等问题。提出一种基于多解码器不确定性感知体系的模型MUA-Net。通过引入专业知识推断模块,将各个专家注释的专业知识水平作为先验知识嵌入编码器和解码器的瓶颈中,以形成包含专家线索的高级语义特征。利用可同时学习多个注释的多解码器结构调节多专家之间的分歧,重构多专家注释过程,并对不确定或分歧区域进行量化。提出一种双分支软注意机制,增强多解码器分割预测的模糊区域,得到最终校准的分割结果。实验结果表明,该模型在RIGA数据集上能以较高的不确定性预测合理的区域,与MRNet模型相比,该模型在视杯分割中的平均精度、Dice系数、交并比分别提升了0.75、0.39、0.41个百分点。     

基于神经网络和信息检索的源代码注释生成

源代码注释生成旨在为源代码生成精确的自然语言注释,帮助开发者更好地理解和维护源代码.传统的研究方法利用信息检索技术来生成源代码摘要,从初始源代码选择相应的词或者改写相似代码段的摘要;最近的研究采用机器翻译的方法,选择编码器-解码器的神经网络模型生成代码段的摘要.现有的注释生成方法主要存在两个问题:一方面,基于神经网络的方法对于代码段中出现的高频词更加友好,但是往往会弱化低频词的处理;另一方面,编程语言是高度结构化的,所以不能简单地将源代码作为序列化文本处理,容易造成上下文结构信息丢失.因此,本文为了解决低频词问题提出了基于检索的神经机器翻译方法,使用训练集中检索到的相似代码段来增强神经网络模型;为了学习代码段的结构化语义信息,本文提出结构化引导的Transformer,该模型通过注意力机制将代码结构信息进行编码.经过实验,结果证明该模型在低频词和结构化语义的处理上对比当下前沿的代码注释生成的深度学习模型具有显著的优势.     

切削刀具综合管理系统的数据采集和存储设计研究*

为建立满足汽车发动机制造企业需求的切削刀具综合管理系统,本文研究了该系统数据库的基础数据采集和存储问题。针对生产现场数据采集困难的问题,设计基于射频识别技术的数据采集方法,提高采集的效率和数据的可靠度。提出刀具与工步对应的存储方法,将刀具信息与生产线相应信息有效衔接。建立系统E-R模型,应用关系规范化理论进行数据存储分析,设计数据表及其相互关系,解决数据库存储的冗余问题。以上述研究成果为基础建立切削刀具数据库,并验证数据库企业实施的可行性。研究成果为建立具有高可靠性和操作性的刀具综合管理系统提供了理论和实践保障。     

基于卷积神经网络的代码注释自动生成方法

自动化代码注释生成技术通过分析源代码的语义信息生成对应的自然语言描述文本,可以帮助开发人员更好地理解程序,降低软件维护的时间成本.大部分已有技术是基于递归神经网络(Recurrent Neural Network,RNN)的编码器和解码器神经网络实现的,但这种方法存在长期依赖问题,即在分析距离较远的代码块时,生成的注释信息的准确性不高.为此,文中提出了一种基于卷积神经网络(Convolutional Neural Network,CNN)的自动化代码注释生成方法来缓解长期依赖问题,以生成更准确的注释信息.具体而言,通过构造基于源代码的CNN和基于AST的CNN来捕获源代码的语义信息.实验结果表明,与DeepCom和Hybrid-DeepCom这两种最新的方法相比,在常用的BLEU和METEOR两种评测指标下,所提方法能更好地生成代码注释,且执行时间更短.     

融合多结构信息的代码注释生成模型

代码注释可以帮助开发人员理解代码的功能和实现方法。代码注释生成模型可以自动识别代码中的关键信息,并生成相关注释,提高代码的可读性和可维护性。现有的代码注释生成模型通常只使用抽象语法树结构信息来表示代码,导致模型生成注释质量不高。提出一种融合多结构信息的代码注释生成模型,该模型在代码抽象语法树的基础上,增加了数据流图结构信息来表示代码。模型使用Transformer的编码器对抽象语法树序列进行编码,捕获代码全局信息。使用图神经网络对数据流图进行特征提取,提供变量之间的计算依赖关系等信息。然后使用跨模态注意力机制融合抽象语法树和数据流2种特征,经过Transformer的解码器生成相应的注释。实验结果表明,与6种主流模型相比,所提出的模型在Java和Python数据集上的BLEU、METEOR和ROUGE-L指标得分均有提高,生成的注释也具有良好的可读性。     

基于结构感知双编码器的代码注释自动生成

在软件开发过程中,性能良好的代码注释工具能够提高开发效率并降低维护成本。部分研究者将代码注释自动生成看作将源代码翻译成自然语言注释的翻译任务,但仅考虑源代码的序列信息而忽略了代码内部的结构特性。为此,在常见端到端翻译模型的基础上,利用代码抽象语法树将源代码的结构信息嵌入到编码器解码器翻译模型中,提出一种基于结构感知的双编码器解码器模型,该模型综合考虑源代码的序列信息与代码内部的结构特性。在真实数据集上的实验结果表明,相比PBMT、Seq2seq模型,该模型的BLEU得分较高,且生成的注释更准确和易读。     

用人工智能算法实现电力系统结线分析

提出了一种可用于电力系统结线分析的基于产生式系统的人工智能算法.该方法以由元件和开关的关系所建立的数据信息和规则为基础,利用元件所包含的信息,用规则来控制信息的流动,根据人工智能的工作原理,运用产生式规则来确定电力系统结线分析的网络结线.通过对一个电力系统模型的分析,表明该方法的可行性及有效性.     

基于UG的采煤机防爆盖板三维标注及数控工艺研究

随着三维CAD技术的发展,产品制造信息在三维模型和二维图纸之间的反复传递,严重影响了产品制造信息的正确、快速传递。基于三维产品制造信息技术,以采煤机防爆盖板为例,在UG NX10.0中建立了防爆盖板的三维模型并进行了三维标注,设计了加工工艺路线,规划了数控加工刀具轨迹,进行了加工仿真,验证了数控加工的正确性,在此基础上完成了采煤机防爆盖板的三维工艺式设计。该研究对基于模型定义的数字化设计与制造研究,实现制造业信息化具有重要意义。     

机床联网项目概述与实施难点

本文通过对机床联网项目的概述、其核心功能(分布式控制网络、生产数据采集、NC程序管理、可视化数据展示)及其与其它应用系统(PDM/CAPP、MES、刀具管理系统)的集成,分析机床联网项目对机械制造企业提高竞争力的重要性     

可重配置DNC系统的设计与实现

为了使DNC系统适应不同配置的制造单元,同时有效地控制和管理制造单元中数控机床的加工过程,该文从DNC系统的控制软件可重配置以及硬件通讯平台可重配置的角度出发,提出了一种可重配置制造单元中数控机床的DNC控制设计方法,同时采用面向对象技术进行DNC系统软件的设计和开发,所开发的系统集NC程序编辑、NC程序管理、DNC控制为一体,能够满足动态单元和可重构制造单元等DNC的控制要求。     

基于GB/T 24734 的三维尺寸快速标注技术

针对当前主流CAD 软件提供的三维标注功能局限的现状,结合制造企业 对复杂零件三维模型快速标注的需求,探讨了基于GB/T 24734 的三维尺寸快速标注技术; 设计并实现了三维尺寸快速标注的算法,详细阐述了其相关的技术。该算法涉及标注面自动 构造、尺寸元素计算、显示处理等一系列技术。其中,标注面构造主要涉及标注面分类、光 标点获取等操作。为了便于尺寸的显示与观察,采用了标注文字的正向显示处理技术。以 CATIA V5 系统为平台,实现了上述算法,开发了快速标注模块,通过实例测试验证了算法 的可行性。     

Estimation of NC machining time using NC block distribution for sculptured surface machining

The estimation of NC machining time is of importance because it provides manufacturing engineers with information to accurately predict the productivity of an NC machine, as well as its production schedule. NC programs contain various machining information, such as tool positions, feed and speed rates, and other machine instructions. Nominal NC machining time can easily be obtained based on the NC program data. Actual machining time, however, cannot simply be found due to the dynamic characteristics of a NC machine controller, such as acceleration and deceleration effect. Hence, this study presents an NC machine time estimation model for machining sculptured surfaces, considering such dynamic characteristics of the machine. The proposed estimation model uses several factors, such as the distribution of NC blocks, angle between the blocks, federates, acceleration and deceleration constants, classifying tool feed rate patterns into four types based on the acceleration and deceleration profile, NC block length, and minimum feed rate. However, there exists an error for the actual machining time due to the lack of the measurement equipment or tools to gauge an exact minimum feed rate. Thus, this paper proposes a machining time estimation model using NC block distributions, lowering down the error caused by the inaccurate minimum feed rate. The proposed machining time estimator performs at around 10% of mean error.     

STEP-NC based high-level machining simulations integrated with CAD/CAPP/CAM

With the development of manufacturing,numerical control(NC) machining simulation has become a modern tool to obtain safe and reliable machining operations.Although some research and commercial software about NC machining simulations is available,most of them is oriented for G&M code.It is a low-level data model for computer numerical control(CNC),which has inherent drawbacks such as incomplete data and lack of accuracy.These limitations hinder the development of a real simulation system.Whereas,standard for the exchange of product data-compliant numerical control(STEP-NC) is a new and high-level data model for CNC.It provides rich information for CNC machine tools,which creates the condition for an informative and real simulation.Therefore,this paper proposes STEP-NC based high-level NC machining simulations solution integrated with computer-aided design/computeraided process planning/computer-aided manufacturing(CAD/CAPP/CAM).It turned out that the research provides a better informed simulation environment and promotes the development of modern manufacturing.     

A French clinical corpus with comprehensive semantic annotations: development of the Medical Entity and Relation LIMSI annOtated Text corpus (MERLOT)

Quality annotated resources are essential for Natural Language Processing. The objective of this work is to present a corpus of clinical narratives in French annotated for linguistic, semantic and structural information, aimed at clinical information extraction. Six annotators contributed to the corpus annotation, using a comprehensive annotation scheme covering 21 entities, 11 attributes and 37 relations. All annotators trained on a small, common portion of the corpus before proceeding independently. An automatic tool was used to produce entity and attribute pre-annotations. About a tenth of the corpus was doubly annotated and annotation differences were resolved in consensus meetings. To ensure annotation consistency throughout the corpus, we devised harmonization tools to automatically identify annotation differences to be addressed to improve the overall corpus quality. The annotation project spanned over 24 months and resulted in a corpus comprising 500 documents (148,476 tokens) annotated with 44,740 entities and 26,478 relations. The average inter-annotator agreement is 0.793 F-measure for entities and 0.789 for relations. The performance of the pre-annotation tool for entities reached 0.814 F-measure when sufficient training data was available. The performance of our entity pre-annotation tool shows the value of the corpus to build and evaluate information extraction methods. In addition, we introduced harmonization methods that further improved the quality of annotations in the corpus.     

A sensor-driven approach to Web-based machining

The objective of this research is to develop a framework named Wise-ShopFloor and the enabling technologies for collaborative manufacturing in a decentralized environment. Particularly, this paper presents our latest development on Web-based and sensor-driven remote machining. Once a product design is given, its process plan and NC codes are generated by using a distributed process planning (DPP) system. The NC codes are then used for remote machining via a standard Web browser and a Java GUI interface running inside the browser. In this paper, the focus is given to the concept, architecture and a prototype implementation of the enabling technology. A case study of a test part machining on a 5-axis milling machine is also completed for testing and validation. It is expected that the developed technology can be applied to design verification via remote machining as well as real part production in a distributed manufacturing environment.     

3-D object decomposition with extended octree model and its application in geometric simulation of NC machining

The paper discusses the development of a prototype solid modeling system based on the extended octree modeling approach and its applications in 3-D NC machining simulation and automatic verification. Along with a simple hierarchical data structure, the extended octree model uses the face boundary information (i) to represent complex objects, (ii) to improve object representational accuracy, and (iii) to accelerate the model updating procedures in a graphic simulation process. The improved representational accuracy makes it possible to carry out automatic NC program verification by generating the machined model of the part through the simulation and comparing it with the designed model of the part. The paper also addresses the issues of model conversions from CSG and B-Rep schemes to the corresponding extended octree models and the issues of carrying out Boolean operations on those extended octree models. A prototype system has been implemented and is integrated with AutoCAD AME solid modeler for object modeling and for NC simulation purpose.