ScienceDirect
科学直通

Procedia CIRP 98 (2021) 517–522
论文集CIRP98（2021）：517–522

www.elsevier.com/locate/procedia

28th CIRP Conference on Life Cycle Engineering
第 28 届CIRP生命周期工程会议

A Sustainable Reverse Engineering Process
可持续的逆向工程流程

Katsuhiro Saiga^a, AMM Sharif Ullah^b*, Akihiko Kubo^b, Tashi^a
KatsuhiroSaiga a，AMMSharifUllah^b*，AkihikoKubo b，Tashi a

^aGraduate School of Engineering, Kitami Institute of Technology, Kitami 090-8507, Japan
^a 北见工业大学 工学研究生院，日本 北见090-8507

^bDivision of Mechanical and Electrical Engineering, Kitami Institute of Technology, Kitami 090-8507, Japan
^b 北见工业大学 机电工程专业，日本 北见090-8507

* Corresponding author. Tel.: +81-157-26-9207; fax: +81-157-26-9207. E-mail address: ullah@mail.kitami-it.ac.jp
*通讯作者。电话：+81-157-26-9207;传真：+81-157-26-9207.邮箱地址：ullah@mail.kitami-it.ac.jp

Abstract
抽象

Reverse engineering mimics critical features of an existing object to create its accurate or enhanced virtual/physical models. It is thus useful in creating the digital footprints of an object. It requires sophisticated devices, computing facilities, and high human skills. Thus, a reverse engineering process becomes sustainable if it is less dependent on sophisticated devices, computations, and human skills. From this point of view, this study presents a sustainable reverse engineering process consisting of five steps. The first step describes a given object using some elementary geometric shapes. The second step represents the elementary geometric shapes using some simulated planner point clouds. The third step creates coordinated point clouds by combining the point clouds of the second step. The fourth step creates a solid model using an off-the-shelf CAD package. The fifth step creates a 3D printed prototype using the solid CAD model. The process is compared with the traditional scanned point cloud-based reverse engineering. The comparison shows that it is free from computational complexity, creates comparatively accurate models, and works very fast. Integrating the presented reverse engineering process with a product life cycle engineering can contribute to sustainable product development.
逆向工程模拟现有对象的关键特征，以创建其准确或增强的虚拟/物理模型。因此，它在创建对象的数字足迹时非常有用。它需要复杂的设备、计算设施和高超的人类技能。因此，如果逆向工程过程较少依赖复杂的设备、计算和人类技能，它就会变得可持续。从这个角度来看，本研究提出了一个由五个步骤组成的可持续逆向工程过程。第一步使用一些基本几何形状来描述给定的对象。第二步使用一些模拟的规划器点云表示基本几何形状。第三步通过组合第二步的点云来创建协调的点云。第四步使用现成的 CAD 软件包创建实体模型。第五步使用实体 CAD 模型创建 3D 打印原型。该流程与传统的基于扫描点云的逆向工程进行了比较。比较表明，它没有计算复杂性，可以创建相对准确的模型，并且运行速度非常快。 将所提出的逆向工程流程与产品生命周期工程相结合，有助于可持续的产品开发。

© 2021 The Authors. Published by Elsevier B.V.
© 2021作者。出版商 Elsevier B.V.

This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
这是一篇在CCBY-NC-ND许可 （https://creativecommons.org/licenses/by-nc-ND/4.0） 下的开放获取文章

Peer-review under responsibility of the scientific committee of the 28th CIRP Conference on Life Cycle Engineering.
由第 28 届CIRP生命周期工程会议科学委员会负责的同行评审。

Keywords: Reverse Engineering; Sustainability; Product Development; Product Life cycle, Human Cognition; Geometric Modeling
关键词：逆向工程;可持续性;产品开发;产品生命周期，人类认知;几何建模

Introduction
介绍

According to the United Nations (UN), sustainability means achieving current needs without jeopardizing the potential of fulfilling future needs [1]. Seventeen goals, known as the Sustainable Development Goals (SDGs), have been introduced to ensure sustainability [2]. Goal 12 (ensure sustainable consumption and production patterns) is highly relevant to the product/system life cycle. Two of the remarkable sub-goals of goal 12 are as follows: 1) By 2030, achieve sustainable management and efficient use of natural resources. 2) By 2030, substantially reduce waste generation through prevention, reduction, recycling, and reuse. Whether or not these sub-goals are achieved can be ensured by measuring the indicators called material, energy, and product efficiencies [3,4]. Nevertheless, other indicators can be considered to make sustainability more meaningful. In this case, an indicator defined as system
根据联合国 （UN） 的说法，可持续性意味着在不危及满足未来需求的潜力的情况下满足当前需求 [1]。为确保可持续性，已经引入了 17 个目标，即可持续发展目标 （SDG） [2]。目标 12（确保可持续的消费和生产模式）与产品/系统生命周期高度相关。目标 12 的两个显著子目标如下：1） 到 2030 年，实现自然资源的可持续管理和有效利用。2）到2030 年，通过预防、减少、回收和再利用，大幅减少废弃物的产生。这些子目标是否实现可以通过测量称为材料、能源和产品效率的指标来确保 [3,4]。尽管如此，可以考虑其他指标来使可持续性更有意义。在这种情况下，定义为system 的指标

efficiency is considered in this study. The description of this indicator is as follows.
本研究考虑了效率。该指标的描述如下。

A set of systems (System X1,…) is needed to carry out the
一套系统（SystemX1,...）执行

activities associated with the life cycle stages, as shown in Fig.
与生命周期阶段相关的活动，如图1 所示。

The systems can work independently or concurrently. However, if the systems are expensive and require highly sophisticated computational procedures and highly skilled human resources, the stakeholders might encounter unwanted complexity. As a result, the activities needed to support the product life cycle stages may not be performed as expected, resulting in unnecessary wastage of time and resources. These problems can be measured quantitatively or qualitatively using an indicator denoted as system efficiency. Therefore, low system efficiency prevents smooth executions of the activities associated with the life cycle stages. How to achieve high system efficiency is thus an important aspect of sustainability or product life cycle. This study addresses this issue.
这些系统可以独立工作，也可以同时工作。但是，如果系统成本高昂，并且需要高度复杂的计算过程和高技能的人力资源，则利益相关者可能会遇到不必要的复杂性。因此，支持产品生命周期阶段所需的活动可能无法按预期执行，从而导致不必要的时间和资源浪费。这些问题可以使用表示为系统效率的指标进行定量或定性测量。因此，系统效率低会阻止与生命周期阶段相关的活动的顺利执行。因此，如何实现高系统效率是可持续性或产品生命周期的一个重要方面。本研究解决了这个问题。

2212-8271 © 2021 The Authors. Published by Elsevier B.V.
2212-8271©2021作者。出版商ElsevierB.V.

This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the 28th CIRP Conference on Life Cycle Engineering. 10.1016/j.procir.2021.01.144
这是一篇在CCBY-NC-ND许可（https://creativecommons.org/licenses/by-nc-nd/4.0）同行评审下的开放获取文章，由第 28 届CIRP会议科学委员会负责关于生命周期工程。10.1016/j.procir.2021.01.144

Particularly, system efficiency from the perspective of reverse engineering [5] is considered, which is directly associated with the stages called strategies, customer needs, concepts, virtual model, selected model, and prototypes (Fig. 1). In “conventional engineering,” multiple concepts of a product are first generated based on the customer needs and strategy of the relevant organization. The concepts are then converted into product models (virtual models). Subsequently, the optimal virtual models are converted into physical models (prototypes) before going for manufacturing. Thus, conventional engineering refers to the pathway “concept-model-object.” Many systems are used to supports this pathway. Sometimes an opposite pathway is selected called reverse engineering [5]. As the naming suggests, reverse engineering refers to the pathway “object-model-concept”. Thus, in reverse engineering, the valuable information extracted from an existing object is utilized to create a model. From the model, the concept underlying the object is extracted. The above description reverse engineering is somewhat a classical description. In reality, reverse engineering takes a somewhat different form. It
特别是，从逆向工程 [5] 的角度考虑了系统效率，它与策略、客户需求、概念、虚拟模型、选定模型和原型等阶段直接相关（图 D）。1）. 在“传统工程”中，首先根据客户需求和相关组织的战略生成产品的多个概念。然后将这些概念转换为产品模型（虚拟模型）。随后，在制造之前，将最佳虚拟模型转换为物理模型（原型）。因此，传统工程是指“概念-模型-对象”路径。许多系统用于支持此通路。有时会选择相反的途径，称为逆向工程[5]。顾名思义，逆向工程是指“对象-模型-概念”路径。因此，在逆向工程中，从现有对象中提取的有价值信息被用于创建模型。从模型中提取对象的基础概念。上述描述逆向工程在某种程度上是一种经典的描述。实际上，逆向工程的形式略有不同。它

mimics some of the critical geometric features of an existing object, creates its accurate or enhanced virtual/physical models, and generates digital footprints of the object, as shown in Fig.
模拟现有物体的一些关键几何特征，创建其精确或增强的虚拟/物理模型，并生成物体的数字足迹，如图 1 所示。

In most cases, scanned point clouds are used as input information, as seen in Fig. 2. A review of real-life reverse engineering is given in the next section. As described in Section 2, conventional reverse engineering requires sophisticated devices and computing facilities. From the viewpoint of system efficiency, a reverse engineering process that is less dependent on sophisticated devices and complex computations exhibits can exhibit high system efficiency and become more sustainable. From the above consideration, the article is written. The rest of this article is organized as follows: Section 2 presents a literature review. Section 3 outlines the steps of the proposed sustainable reverse engineering process. Section 4 presents two case studies implementing the proposed reverse engineering process. T This section also compares results with the traditional scanned point cloud-based reverse engineering, exhibiting the proposed reverse engineering’s efficacy. Section
在大多数情况下，扫描的点云用作输入信息，如图 1 所示。 2.下一节将对现实生活中的逆向工程进行回顾。如第 2 节所述，传统的逆向工程需要复杂的设备和计算设施。从系统效率的角度来看，对复杂设备和复杂计算的依赖较少的逆向工程过程可以表现出较高的系统效率，并成为更可持续。从以上考虑出发，写了这篇文章。本文的其余部分组织如下：第 2 部分介绍了文献综述。第 3 节概述了拟议的可持续逆向工程流程的步骤。第 4 节介绍了两个实施拟议逆向工程过程的案例研究。 本节还将结果与传统的基于扫描点云的逆向工程进行了比较，展示了所提出的逆向工程的有效性。部分

5 concludes this article.
5本文总结道。

Strategies
策略

Customers Needs
客户需求

Concepts Virtual Models Selected Solutions Prototypes
概念虚拟模型选定的解决方案原型

Strategies Customers Needs Concepts
策略客户需求概念

Selected Solutions Prototypes
选定的解决方案原型

Use
用

Service
服务

System X₁ System X
System X系统 X₂

…

System Xn1 System X
System X系统 Xn

Service
服务

Use
用

Supply
供应

Manufacturing
制造业

Disposal Recycle Landfill
处置回收垃圾填埋场

Prototype
原型

Enhanced model
增强模型

Point cloud
点云

limitations. According to this survey, a reverse engineering process depends on the solid model (virtual model) reconstruction process using mostly off-the-shelf CAD packages.
局限性。根据这项调查，逆向工程过程取决于实体模型（虚拟模型）重建过程，该过程主要使用现成的 CAD 软件包。 This
这 process
过程 starts with
开头为 point
点 cloud data acquisition and terminates at virtual model construction. The typical intermediate steps are preprocessing, segmentation, feature classification, and mesh/surface fitting. The remarkable thing is
云数据采集，并在虚拟模型构建时终止。典型的中间步骤是预处理、分割、特征分类和网格/表面拟合。值得注意的是 that
那 the
这 laser-scanned
激光扫描 point
点 clouds
云 exhibit
展览 noise
噪声 and
和 outliers. As a result, preprocessing and post-processing are required before extracting the geometric features. Numerous authors delved
异常。因此，在提取几何特征之前，需要进行预处理和后处理。众多作者深入研究 into
到 the
这 preprocessing
预处理 and
和 post-processing
后处理 of
之 a scanned point cloud. For the relatively small-size point clouds, one of the
扫描的点云。对于相对较小的点云， well-known
有名 preprocessing
预处理 techniques
技术 is
是 the
这 iterative
迭 代 closest point
最近点 algorithm
算法 [7, 8]. Many
多 authors
作者 have
有 applied
应用的 it
它 in
在 real-life reverse
现实生活中的反向 engineering.
工程。 For
为 example,
例 Guan
官 and
和 Gu
古 [9] used
使用 both
双

Fig. 2. Conventional reveres engineering.
无花果。2.常规崇尚工程。

Related Work
相关工作

This section reviews the relevant literature describing various types of real-life reverse engineering techniques and their application areas.
本节回顾了描述各种现实生活中的逆向工程技术及其应用领域的相关文献。

Buonamici et al. [6] surveyed the scanned point cloud-based reverse engineering (Fig. 2) and reported its advantages and
Buonamici等人[6] 调查了扫描的基于点云的逆向工程（图 .2）并报告了它的优势和

iterative closest point algorithm and chord deviation calculation to eliminate the noisy points before surface reconstruction of the bowl and blade of a ship. However, the relatively large-size point clouds are treated differently. Particularly, feature extraction from a large-spatial point cloud has been an issue for creating digital footprints of buildings and relevant systems. For example, Araújo and Oliveira [10] developed a fast cylinder-detection technique from a large point cloud (point cloud of a chemical plant). The detection
迭代最近点算法和弦偏差计算，以消除船舶碗和叶片表面重建之前的噪声点。但是，相对较大的点云的处理方式不同。特别是，从大空间点云中提取特征一直是创建建筑物和相关系统的数字足迹的一个问题。例如，Araújo 和 Oliveira [10] 从大型点云（化工厂的点云）开发了一种快速圆柱体检测技术。检测

technique employs a linear-time circle-detection algorithm. The algorithm projects the point cloud onto a set of uniformly- distributed directions defined over the unit hemisphere. It calculates points whose normal vectors are approximately perpendicular to a given projection and refines these directions. The extracted cylindrical surfaces are obtained by fitting a cylinder to each connected component that passes a validity test. Moritani et al. [11] articulated a method to preprocess a laser- scanned point cloud of an arbitrary space to extract the digital footprint of a piping system for renovating it (piping system). The remarkable thing is that the simulated point cloud of the core feature (a cylinder) must interact with the scanned point cloud while removing the noise and outliers. This is achieved by employing a sophisticated algorithm, namely, iteratively reweighted least-squares method. Massafra et al. [12] developed a method of dealing with a large point cloud for feature extraction. This method first divides a massive point cloud into small feature-based points cloud for constructing the digital footprints of wooden trusses of a historical building. Finally, parametric modeling was added to refine the objects. Other computer-aided engineering techniques can aid the refinement process. This way, more effective building information modeling systems can be constructed. In order to develop a building facility’s digital model, a semi-automatic method was developed by Zeng et al. [13]. In this method, first, a pre-trained deep neural network extracts a 50-dimensional feature vector for each point. This helps segment the point clouds in clusters where the region-growing algorithms are used. The user-selected visualized features are then used as examples to run the peak-finding algorithm to determine positive matches.
技术采用线性时间圈检测算法。该算法将点云投影到在单位半球上定义的一组均匀分布的方向上。它计算法向矢量大致垂直于给定投影的点，并优化这些方向。提取的圆柱面是通过将圆柱体拟合到通过有效性测试的每个连通分量来获得的。Moritani等[11]提出了一种方法，可以对任意空间的激光扫描点云进行预处理，以提取管道系统的数字足迹，用于翻新管道系统（管道系统）。值得注意的是，核心特征（圆柱体）的模拟点云必须与扫描的点云交互，同时去除噪声和异常值。这是通过采用一种复杂的算法来实现的，即迭代重新加权的最小二乘法。Massafra等人[12]开发了一种处理大型点云以进行特征提取的方法。该方法首先将一个巨大的点云划分为基于特征的小点云，用于构建历史建筑木桁架的数字足迹。最后，添加了参数化建模以优化对象。其他计算机辅助工程技术可以帮助优化过程。这样，可以构建更有效的建筑信息模型系统。为了开发建筑设施的数字模型，Zeng等人[13] 开发了一种半自动方法。在这种方法中，首先，一个预先训练的深度神经网络为每个点提取一个50 维的特征向量。这有助于对使用区域增长算法的集群中的点云进行分段。然后，将用户选择的可视化特征用作示例，以运行峰值查找算法以确定正匹配项。

In addition to scanned point clouds, imaging plays a vital role in reverse engineering. For example, Mengoni and Leopardi [14] summarized the performances of the hardware and software tools used in both active sensor or range-based technology (3D laser scanner) and Structure from Motion (SfM) based technology reverse engineering applied to digital documentation of archeological heritage. The former one directly acquires the coordinates of the exposed surfaces of an object. These coordinates are converted into a 3D solid model after performing the necessary pre- and post-processing, as mentioned before. On the other hand, the latter (SfM) combines a series of photos of an object and produces its 3D solid model. SfM is good at constructing an object's facsimile. When the object is scaled, high-resolution point clouds obtained by using a 3D scanner must be integrated with the SfM technique to ensure the accuracy of the object. In addition to SfM, other imaging techniques can be used for reverse engineering. For example, multiple spin-image can be used to reconstruct a voxel-based digital model of an object, as shown by Nanya et al. [15]. The voxels are generated from the images. This method requires a manual check on some edges, ensuring the targeted edge is either inside or outside the desired region of the image. This results in the elimination of redundant voxels. Reconstruction of silhouetted concave regions and extremely thin segments are two of the main challenges of the method. To overcome these challenges, the hardware/software used in data acquisition (imaging) needs improvement.
除了扫描的点云外，成像在逆向工程中也起着至关重要的作用。例如，Mengoni 和 Leopardi [14] 总结了用于主动传感器或基于距离的技术（3D 激光扫描仪）和基于运动结构 （SfM） 的技术逆向工程应用于考古遗产数字记录中使用的硬件和软件工具的性能。前者直接获取物体裸露表面的坐标。如前所述，在执行必要的前处理和后处理后，这些坐标将转换为 3D 实体模型。另一方面，后者 （SfM） 结合了对象的一系列照片并生成其3D实体模型。SfM 擅长构造对象的传真。在对物体进行缩放时，使用3D 扫描仪获得的高分辨率点云必须与 SfM 技术相结合，以确保物体的精度。除了 SfM 之外，其他成像技术也可用于逆向工程。例如，多个自旋图像可用于重建物体的基于体素的数字模型，如 Nanya 等人 [15] 所示。体素是从图像生成的。此方法需要手动检查某些边缘，确保目标边缘位于图像所需区域的内部或外部。这将消除冗余体素。 重建轮廓凹面区域和极薄的段是该方法的两个主要挑战。为了克服这些挑战，用于数据采集（成像）的硬件/软件需要改进。

As far as quality control is concerned, reverse engineering has been very instrumental. Zachos et al. [16] showed how to use the point cloud of a manufactured turbine blade to determine its accuracy compared to its design. Segreto et al.
就质量控制而言，逆向工程一直发挥着非常重要的作用。Zachos等[16]展示了如何使用制造的涡轮叶片的点云来确定其与设计相比的精度。Segreto等人。

[17] developed a point cloud-based reverse engineering technique useful for quality control of aerospace components made of carbon fiber-reinforced polymer composites. The study showed that the comparison between 3D digital models of the manufactured part and its design requires dedicated user- interfaces. Urata et al. [18] developed a reverse engineering method to produced high surface finished casted products. In this method, first, an existing casted part is scanned to prepare its surface mesh data. The data are processed using a customized geometric modeling technique to separate the as- cast and machined surfaces. After that, the boundary is modified to create a digital model of casting. This way, high- quality mass production using casting can be achieved. Yang et al. [19] used reverse engineering in welding quality control. They have developed a weighted neighborhood search algorithm to delete points from the point cloud of a welded part and extract the points representing a subtle feature (e.g., welded sections of a spherical object). The point cloud elimination process requires a complex computational procedure based on a multi-threshold weighting adjustment calculation customized for the targeted features.
[17] 开发了一种基于点云的逆向工程技术，可用于对碳纤维增强聚合物复合材料制成的航空航天部件进行质量控制。研究表明，制造零件的 3D 数字模型与其设计之间的比较需要专用的用户界面。Urata等[18]开发了一种逆向工程方法来生产高表面精加工的铸造产品。该方法首先扫描现有铸造件，准备其表面网格数据。使用定制的几何建模技术对数据进行处理，以分离铸造表面和机加工表面。之后，修改边界以创建铸造的数字模型。这样，就可以实现使用铸造的高质量批量生产。Yanget al. [19] 在焊接质量控制中使用了逆向工程。他们开发了一种加权邻域搜索算法，可以从焊接零件的点云中删除点，并提取表示细微特征的点（例如，球形物体的焊接截面）。点云消除过程需要一个复杂的计算过程，该过程基于为目标特征定制的多阈值加权调整计算。

When an object is not available, sketches can be used to carry out reverse engineering, as sketches are a natural and intuitive means to express concepts [20]. In this case, a sketch must be converted into an off-the-shelf CAD compatible solid model. Different types of machine learning algorithms are used to reverse engineer an object from its sketch [20]. In these algorithms, the topological rule-bases of some selected vertex- edge pairs are used. These rules must establish the topological relationships among the edges and vertices at the rough intersections of a sketch. The rules thus help interpret a rough sketch in terms of well-formed geometrical entities. The rules are often user-defined and categorized in terms of features (e.g., rectangular box, cylindrical shape, through holes, and alike). For example, Tanaka et al. [21] developed sex sets of rules to develop a system. The system converts a rough sketch (composed of curved and straight edges) into a CAD package compatible solid model.
当一个物体不可用时，草图可以用来进行逆向工程，因为草图是表达概念的一种自然而直观的方式[20]。在这种情况下，必须将草图转换为现成的 CAD 兼容实体模型。不同类型的机器学习算法用于从草图对对象进行逆向工程 [20]。在这些算法中，使用了一些选定的顶点边缘对的拓扑规则库。这些规则必须在草图的粗略交点处建立边和顶点之间的拓扑关系。因此，这些规则有助于根据格式正确的几何实体来解释粗略草图。这些规则通常是用户定义的，并根据特征（例如，矩形框、圆柱形、通孔等）进行分类。例如，Tanaka et al. [21] 开发了性规则集来开发一个系统。系统将粗略草图（由弯曲和直边组成）转换为与 CAD 软件包兼容的实体模型。

In addition to scanned point clouds, images, and sketches, there is another type of reverse engineering where analytical points are used to mimic an object's important features. The analytical point clouds can be created in different ways. For example, Montlahuc et al. [22] presented an analytical point cloud-based reverse engineering method. The method analytically creates as-scanned point clouds of industrial assembly models. It first generates a watertight triangle mesh wrapping the relevant CAD model. The method finally converts the mesh into a point cloud. Since the method does not use a laser scanner to obtain the point cloud, it is free from computational complexities (redundant and noisily points elimination, surface construction from the point cloud, and alike). This does not mean that it is free from all sorts of computational complexity. Notably, how to control the distribution of points, remove the hidden points, and adjust the misalignments are some of the computational challenges of the
除了扫描的点云、图像和草图外，还有另一种类型的逆向工程，其中分析点用于模拟对象的重要特征。分析点云可以通过不同的方式创建。例如，Montlahuc et al. [22] 提出了一种基于分析点云的逆向工程方法。该方法以分析方式创建工业装配模型的扫描点云。它首先生成一个包裹相关 CAD 模型的水密三角形网格。该方法最终将网格转换为点云。由于该方法不使用激光扫描仪来获取点云，因此它没有计算复杂性（冗余和嘈杂的点消除、点云的表面构造等）。这并不意味着它没有各种计算复杂性。值得注意的是，如何控制点的分布、去除隐藏点和调整错位是

method. Tashi et al. [23-25] developed an analytical point cloud-based reverse engineering. In this case, a recursive process is used to create a planner point cloud. The process is originated from work done by Ullah et al. [26]. It is effective in reverse engineering 2.5 and 3D shapes. This method helps develop a new breed of reverse engineering defined as human cognitive reverse engineering [27].
方法。Tashi等[23-25]开发了一种基于分析点云的逆向工程。在这种情况下，使用递归过程来创建规划器点云。该过程起源于Ullah等人所做的工作[26]。它在逆向工程 2.5 和 3D 形状中很有效。这种方法有助于开发一种新型的逆向工程，即人类认知逆向工程[27]。

Method
方法

The proposed sustainable reverse engineering process is schematically illustrated in Fig. 3. The process integrates humans and machines to achieve its goal and discourages scanned point clouds’ direct involvement. Rather it employs planner point clouds simulated by analytical approaches.
所提出的可持续逆向工程过程如图 3 所示。该过程将人和机器集成在一起以实现其目标，并阻止扫描的点云直接参与。相反，它采用通过分析方法模拟的规划器点云。

Elementary and
Elementary和

coordinated point cloud denoted as 4. The coordinated point cloud maintains at least the C⁰ continuity among its constituents. Thus, after completing steps 2 and 3, the reverse engineering process is populated with both elementary and coordinated point clouds.
协调点云表示为 4。协调的点云在其组成部分之间至少保持 C⁰ 连续性。因此，在完成步骤 2 和 3 后，逆向工程过程将同时填充基本点云和协调点云。

The fourth step creates a solid model of the elementary and coordinated point clouds using an off-the-shelf CAD package. For example, as shown in Fig. 3, the point clouds of 3 and 4 are converted into two sloid models denoted as 5 and 6, respectively, using a commercially available CAD package. (The name is not disclosed to avoid commerciality.) It is worth mentioning that most commercially available CAD packages nowadays offer a function to directly import the coordinates of the points comprising a point cloud. The imported point clouds can be converted into (most likely) closed curves/lines before performing solid modeling using some predefined functions, e.g., rotation, extrusion, loft, sweep, shell, and alike, see
第四步使用现成的 CAD 软件包创建基本和协调点云的实体模型。例如，如图1 所示。3、使用市售的CAD软件包将3和4的点云转换为两个SLOID模型，分别表示为5和6。（该为避免商业性，不披露姓名。值得一提的是，现在大多数市售的 CAD 软件包都提供了直接导入构成点云的点坐标的功能。导入的点云可以转换为（很可能）闭合的曲线/线条，然后再使用一些预定义的功能进行实体建模，例如旋转、拉伸、放样、扫掠、壳等，请参阅

Object
对象

Prototype
原型

High-level Description
概要描述

Plate Disk
板盘

3D Printing
3D打印

Coordinated Point Clouds
协调点云

Solid Modeling using CAD
使用CAD 进行实体建模

https://web.mit.edu/2.972/www/solid_modeling.html). The solid models of the elementary point clouds and coordinated point clouds can be added to create the object’s virtual model to be reverse-engineered. For example, as shown in Fig. 3, solid model 7 is the virtual model of the real object created by adding solid models 5 and 6.
https://web.mit.edu/2.972/www/solid_modeling.html）。可以添加基本点云和协调点云的实体模型，以创建要进行逆向工程的对象的虚拟模型。例如，如图1 所示。3、实体模型7是通过添加实体模型 5 和 6 创建的真实对象的虚拟模型。

The last step is 3D printing. In this step, a suitable additive manufacturing process can create the prototype using the virtual model’s information. For example, the prototype shown in Fig. 3 is created using a commercially available 3D printer where the input information is model 7 (Fig. 3).
最后一步是 3D 打印。在此步骤中，合适的增材制造工艺可以使用虚拟模型的信息创建原型。例如，图 3 所示的原型是使用市售的 3D 打印机创建的，其中输入信息为模型 7（图 3）。

Fig. 3. The proposed sustainable reverse engineering method.
无花果。3.提出的可持续逆向工程方法。

As seen in Fig. 3, the process consists of five major steps. In synopsis, the first step describes a given object using some elementary geometric shapes. The second step represents the elementary geometric shapes using some simulated planner point clouds. The third step creates coordinated point clouds by combining some of the elementary point clouds. The fourth step creates a virtual model of the object using an off-the-shelf CAD package where the input information is some selected elementary and coordinated point clouds. The fifth step creates a 3D printed prototype using the virtual model. The descriptions of the steps are as follows:
如图 1 所示。3、该过程包括五个主要步骤。在概要中，第一步使用一些基本的几何形状来描述给定的对象。第二步使用一些模拟的规划器点云表示基本几何形状。第三步通过组合一些基本点云来创建协调的点云。第四步使用现成的 CAD 软件包创建对象的虚拟模型，其中输入信息是一些选定的基本和协调点云。第五步使用虚拟模型创建 3D 打印原型。步骤说明如下：

As mentioned above, the first step creates a high-level description of the object to be reverse-engineered. The output of this step depends on the individual who describes the object using some linguistic expressions. For example, according to the perception of an individual familiar with the basic
如上所述，第一步为要进行逆向工程的对象创建高级描述。此步骤的输出取决于使用某些语言表达式描述对象的个人。例如，根据熟悉基本

Results and Discussions
结果与讨论

This section presents two case studies to elucidate the efficacy of the proposed sustainable reverse engineering.
本节提供了两个案例研究来阐明拟议的可持续逆向工程的有效性。

The first case study deals with the reverse engineering of the object shown in Fig. 3. The reverse engineering of the object using the proposed process is already presented in the previous section (Fig. 3). This section shows its comparison with conventional reverse engineering. The results are shown in Fig.
第一个案例研究涉及图 3 所示对象的逆向工程。使用建议的流程对对象进行逆向工程已在上一节中介绍（图 3）。本节展示了它与传统逆向工程的比较。结果如图 1 所示。

As seen in Fig. 4, when the object is scanned to create its point cloud, it exhibits noise and outliers at the intersecting edges. As a result, the virtual model is quite inaccurate. This is reflected in the 3D printed prototype. On the other hand, the virtual model and prototype created using the presented sustainable reverse engineering process are free from shape inaccuracies (Fig. 3).
如图 4 所示，当扫描对象以创建其点云时，它会在相交边缘显示噪声和异常值。因此，虚拟模型非常不准确。这反映在 3D 打印原型中。另一方面，使用所提出的可持续逆向工程流程创建的虚拟模型和原型没有形状不准确（图 3）。

geometrical shapes, the object shown in Fig. 3 consists of three elementary shapes named cylinder, plate, and disk.
几何形状，对象如图 1 所示。3由三个基本形状组成，分别是圆柱形、板形和圆盘形。

The second and third steps create the elementary and coordinated point clouds, respectively, representing the planner features of the shapes. For creating the elementary point clouds, the algorithm shown in [23-26] can be used. For example, the three elementary point clouds shown in Fig. 3 are created using the algorithm shown in [23-26]. The point clouds denoted as 1, 2, and 3 represent the cross-sectional areas of the elementary shapes called cylinder, plate, and disk, respectively. For the sake of solid modeling in step 4, some point clouds can be integrated to create the coordinated point clouds. For example,
第二步和第三步分别创建基本点云和协调点云，表示形状的计划器特征。为了创建基本点云，可以使用 [23-26] 中所示的算法。例如，图 3 中所示的三个基本点云。3个是使用 [23-26] 中所示的算法创建的。表示为1、2 和 3 的点云分别表示称为圆柱体、板和圆盘的基本形状的横截面积。为了在步骤 4 中进行实体建模，可以集成一些点云来创建协调的点云。例如，

Object
对象

Prototype
原型

Scanning Scanned Point Cloud
扫描扫描的点云

in Fig. 3, the point clouds 1 and 2 are integrated to create the
在图 .3、将点云1和2集成在一起，创建

Fig. 4. Conventional reverse engineering of the object shown in Fig. 3.
无花果。4.图 1 所示物体的常规逆向工程。3.

The other case study is shown in Figs. 5 and 6. In particular, Fig. 5 shows the result when conventional reverse engineering is applied. In this case, the object (relatively complex compared to the object shown in Fig. 3) is scanned using a laser scanner. The point clouds result in inaccurate virtual models, and the inaccuracy persists even when the noise and outliers are removed. The degree of inaccuracy this time (Fig. 5) is severe than the previous case (Fig. 4). For this reason, it is not printed.
另一个案例研究如图 1 所示。5和6.特别是，图 .图 5显示了应用传统逆向工程时的结果。在这种情况下，对象（与图 1 中所示的对象相比相对复杂）。3）使用激光扫描仪进行扫描。点云会导致虚拟模型不准确，即使去除噪声和异常值，不准确仍然存在。这次的不准确程度（图 5）比前一次严重（图 5）。4）.因此，它没有被打印出来。

rotated to create the last coordinated point cloud denoted as point cloud 9. The point clouds 8 and 9 can be imported to an off-the-shelf CAD package to perform geometric modeling. The point cloud to solid model (model 11) conversion process is shown by entity 10 where two planar boundaries (point clouds 8 and 9) separated by the desired height are extruded to create the solid model (model 11). Thus, the solid modeling this time creates models 11 and 12. These two models can be added to create the virtual model (denoted as 13). The virtual model
旋转以创建最后一个协调的点云，表示为点云 9。点云 8 和 9 可以导入到现成的 CAD 软件包中，以执行几何建模。点云到实体模型（模型 11）的转换过程由实体 10 显示，其中两个平面边界（点云8和9）被拉伸，由所需高度分隔以创建实体模型（模型11）。因此，这次的实体建模创建了模型11和12。可以添加这两个模型来创建虚拟模型（表示为13）。虚拟模型

Object
对象

Scanned Point Cloud
扫描的点云

can be printed using an available 3D printer for manufacturing the prototype, as shown in Fig. 6. Therefore, the presented reverse engineering produces better results than the scanned point cloud-based conventional reverse engineering.
可以使用可用的3D打印机打印以制造原型，如图 6 所示。因此，所提出的逆向工程比基于扫描点云的常规逆向工程产生了更好的结果。

To be more specific, a quantitative comparison between the conventional and proposed reverse engineering is given in Table 1. Five parameters are used for the sake of comparison, namely, scanning time (min), size of the point clouds (number of points), point cloud editing time (min), geometric modeling time (min), and shape accuracy. The scanning time measures time to scan an object using a commercially available scanner under the standard scanning conditions. For the object shown in Figs. 5 and 6, the scanning time is about 11 minutes. The point clouds’ size for conventional reverse engineering (Fig. 4) is 15164 points, whereas it is only 980 points (models 8 and 9 in Fig. 5) for the proposed method. Point cloud editing time
更具体地说，表 1 给出了传统和拟议逆向工程之间的定量比较。为了进行比较，使用了 5 个参数，即扫描时间 （min）、点云大小 （点数）、点云编辑时间 （min）、几何建模时间 （min） 和形状精度。扫描时间测量在标准扫描条件下使用市售扫描仪扫描对象的时间。对于图 5 和图 6 所示的物体，扫描时间约为 11 分钟。传统逆向工程的点云大小（图 D）。4） 是 15164 点，而它只有980 点（图 8 中的模型 8 和 9）。5）对于建议的方法。点云编辑时间

Fig. 5. Conventional reverse engineering of relatively complex object.
无花果。5.相对复杂对象的常规逆向工程。

When the object shown in Fig. 5 reverse engineered using the proposed reverse engineering process, the results improve, as shown in Fig. 6. As seen in Fig. 6, the object’s high-level description (the object consists of the shapes called cylinder, plate, and disk) remains the same, but its elementary and coordinated point clouds are somewhat different. Particularly, point clouds 1, 2, and 3, representing a circular arc, rectangular boundary, and circle, respectively, constitute the elementary point clouds. The coordinated point clouds 5, 6, and 7 are created by rotating the other coordinated point cloud, point cloud 4, preserving the object’s symmetry. The coordinated point cloud is created, adding the elementary point clouds 1 and 2 and subtracting one of the lines in the point cloud 2. The coordinated point clouds 4, 5, 6, and 7 can be added, maintaining C0 continuity to create another coordinated point cloud denoted as 8, as shown in Fig. 6. The point cloud can be
当图 5 所示的对象使用建议的逆向工程过程进行逆向工程时，结果会得到改善，如图 6 所示。如图 6 所示，对象的高级描述（对象由称为圆柱体、板和圆盘的形状组成）保持不变，但其基本点云和协调点云略有不同。具体而言，点云1、2 和3 分别表示圆弧、矩形边界和圆，构成了基本点云。协调的点云 5、6 和 7 是通过旋转另一个协调的点云（点云 4）来创建的，同时保持对象的对称性。创建协调点云，将基本点云1 和 2 相加，并减去点云 2 中的一条线。可以添加协调的点云 4、5、6 和 7，保持 C0 连续性以创建另一个表示为8 的协调点云，如图8 所示。6.点云可以是

means removing noise and outliers using the standard functions available to a scanning-based revere engineering system. In contrast, for the proposed method, it is the time necessary to complete the operations in the boundary “Elementary and Coordinated point clouds” (Fig. 6). For the conventional method, it is about 30 minutes, whereas, for the proposed method, it is about 10 minutes. After performing point cloud editing under standard conditions, geometric modeling time for conventional reverse engineering means the time needed to create a solid CAD model. On the other hand, the proposed method, the geometric modeling time means the time needed to perform geometric modeling in the boundary “Solid Modeling using CAD” (Fig. 6). Geometric modeling time for conventional reverse engineering is about 1 minute, whereas it is about 5 minutes for the proposed method. The proposed method outperforms the conventional one (compare the virtual models in Figs. 5 and 6) regarding shape accuracy. In synopsis, the proposed reverse engineering method’s system efficiency is much better than that of the conversational one.
意味着使用基于扫描的 Revere 工程系统可用的标准功能去除噪声和异常值。相比之下，对于所提出的方法，它是在边界 “基本点云和协调点云” 中完成操作所需的时间（图 6）。对于传统方法，大约需要 30 分钟，而对于建议的方法，大约需要 10 分钟。在标准条件下执行点云编辑后，传统逆向工程的几何建模时间是指创建实体 CAD 模型所需的时间。另一方面，所提出的方法，几何建模时间是指在边界 “Solid Modeling using CAD” 中执行几何建模所需的时间（图 6）。传统逆向工程的几何建模时间约为1min，而所提方法的几何建模时间约为 5 min。所提出的方法优于传统方法（比较图 .5和6）关于形状精度。概括地说，所提出的逆向工程方法的系统效率远优于对话方法。

High-level
高级

Elementary and Coordinated point clouds
基本点云和协调点云

Fig. 6. The other application of the proposed sustainable reverse engineering.
无花果。6.提出的可持续逆向工程的另一项应用。

Table 1. Comparison between conventional and proposed reverse engineering.
表1.传统和拟议逆向工程之间的比较。

Concluding Remarks
结束语

The activities associated with the early stages of a product life cycle (i.e., the stages of strategies, customer needs, concepts, virtual model, optimal model, and prototypes) can be performed without sacrificing system efficiency if the proposed reverse engineering is used. The method is free from computational and systemic complexity compared to the scanned point cloud-based conversational reverse engineering. The next phase of this study will study more complex objects having a complex twisted radius of curvature.
如果使用建议的逆向工程，则可以在不牺牲系统效率的情况下执行与产品生命周期早期阶段相关的活动（即战略、客户需求、概念、虚拟模型、最佳模型和原型的阶段）。与基于扫描点云的对话式逆向工程相比，该方法没有计算和系统复杂性。本研究的下一阶段将研究具有复杂扭曲曲率半径的更复杂的物体。

References
引用

Report of the World Commission on Environment and Development (1987) Annex to general assembly document A/42/427. Nairobi, Kenya. The United Nations, New York.
世界环境与发展委员会报告（1987 年），大会文件 A/42/427 附件。肯尼亚内罗毕。联合国，纽约。

Department of Economic and Social Affairs, The United Nations. https://sdgs.un.org/ (last accessed 2020.08.26)
联合国经济和社会事务部。https://sdgs.un.org/（最后访问时间 2020 年 8 月 26 日）

Ullah AMMS, Fuji A, Kubo A, Tamaki, Jun’ichi. Analyzing the Sustainability of Bimetallic Components. International Journal of Automation Technology. 2014;8(5):745-53. doi: 10.20965/ijat.2014.p0745.
Ullah AMMS， Fuji A， Kubo A， Tamaki， Jun'ichi.分析双金属部件的可持续性。国际自动化技术杂志。2014 年;8（5）：745-53。doi：10.20965/ijat.2014.p0745.

Ullah AMMS, Hashimoto H, Kubo A, Tamaki J. Sustainability analysis of rapid prototyping: material/resource and process perspectives. International Journal of Sustainable Manufacturing. 2013;3(1):20-36. doi: 10.1504/ijsm.2013.058640.
UllahAMMS，Hashimoto H，Kubo A，Tamaki J.快速原型制造的可持续性分析：材料/资源和工艺视角。国际可持续制造杂志。2013 年;3（1）：20-36。doi：10.1504/ijsm.2013.058640.

Várady T, Martin RR, Cox J. Reverse engineering of geometric models— an introduction. Computer-Aided Design. 1997;29(4):255-68. doi: 10.1016/S0010-4485(96)00054-1.
VáradyT，Martin RR，Cox J.几何模型的逆向工程——简介。计算机辅助设计。1997;29(4):255-68.doi：10.1016/S0010-4485（96）00054-1.

Buonamici F, Carfagni M, Furferi R, Governi L, Lapini A, Volpe Y. Reverse engineering modeling methods and tools: a survey. Computer- Aided Design and Applications. 2018;15(3):443-64. doi: 10.1080/16864360.2017.1397894.
Buonamici F， Carfagni M， Furferi R， Governi L， Lapini A， Volpe Y.逆向工程建模方法和工具：一项调查。计算机辅助设计与应用。2018 年;15(3):443-64.doi：10.1080/16864360.2017.1397894.

Besl PJ, McKay ND. A method for registration of 3-D shapes. IEEE Transactions on Pattern Analysis and Machine Intelligence. 1992;14(2):239-56. doi: 10.1109/34.121791.
Besl PJ，麦凯 ND。一种用于对准 3-D 形状的方法。IEEETransactions on Pattern Analysis and Machine Intelligence（IEEE 模式分析和机器智能汇刊）。1992 年;14(2):239-56.doi：10.1109/34.121791。

Xu F, Zhao X, Hagiwara I. A Study on Automatic Registration in Reverse Engineering(Mechanical Systems). Transactions of the Japan Society of Mechanical Engineers Series C. 2010;76(771):2861-9. doi: 10.1299/kikaic.76.2861.
徐峰，赵X，Hagiwara I.逆向工程（机械系统）自动注册研究.日本机械工程师学会汇刊C. 系列 2010;76(771):2861-9.doi：10.1299/kikaic.76.2861.

Guan G, Wen-wen G. Reconstuction of Propeller and Complex Ship Hull Surface based on Reverse Engineering. Journal of Marine Science and Technology. 2019;27(6):498-504. doi: 10.6119/JMST.201912_27(6).0002.
Guan G，温-温 G. 基于逆向工程的螺旋桨和复杂船体表面重构.海洋科学与技术杂志.2019 年;27（6）：498-504。doi：10.6119/JMST.201912_27（6）.0002.

Araújo AMC, Oliveira MM. Connectivity-based cylinder detection in unorganized point clouds. Pattern Recognition. 2020;100:107161. doi: 10.1016/j.patcog.2019.107161.
Araújo AMC， Oliveira MM. 无组织点云中基于连接的圆柱体检测。模式识别。2020;100:107161.doi：10.1016/j.patcog.2019.107161.

Moritani R, Kanai S, Date H, Watanabe M, Nakano T, Yamauchi Y. Cylinder-based simultaneous registration and model fitting of laser- scanned point clouds for accurate as-built modeling of piping system. Computer-Aided Design and Applications. 2018;15(5):720-33. doi: 10.1080/16864360.2018.1441239.
森谷 R、金井 S、日期 H、渡边 M、中野 T、山内 Y。基于圆柱体的激光扫描点云同步配准和模型拟合，用于管道系统的精确竣工建模。计算机辅助设计与应用。2018 年;15(5):720-33.doi：10.1080/16864360.2018.1441239.

Massafra A, Prati D, Predari G, Gulli R. Wooden Truss Analysis, Preservation Strategies, and Digital Documentation through Parametric
Massafra A， Prati D， Predari G， Gulli R. 木制桁架分析、保存策略和通过参数化进行数字记录

3D Modeling and HBIM Workflow. Sustainability. 2020;12(12):4975. doi: 10.3390/su12124975.
3D 建模和 HBIM 工作流程。可持续性。2020;12（12）：4975。doi：10.3390/su12124975.

Zeng S, Chen J, Cho YK. User exemplar-based building element retrieval from raw point clouds using deep point-level features. Automation in Construction. 2020;114:103159. doi:10.1016/j.autcon.2020.103159. 8
曾S，陈J，Cho YK.使用深度点级特征从原始点云中检索基于用户示例的建筑元素。建筑自动化。2020;114：103159. doi：10.1016/j.autcon.2020.103159. 8

Mengoni M, Leopardi A. An Exploratory Study on the Application of Reverse Engineering in the Field of Small Archaeological Artefacts. Computer-Aided Design & Applications, 2019; 16(6):1209-26. doi: 10.14733/cadaps.2019.1209-1226.
门戈尼 M， 莱奥帕尔迪 A.逆向工程在小型考古文物领域的应用的探索性研究。计算机辅助设计与应用， 2019; 16（6）：1209-26. doi：10.14733/cadaps.2019.1209-1226.

Nanya T, Yoshihara H, Maekawa T. Reconstruction of Complete 3D Models by Voxel Integration. Journal of Advanced Mechanical Design, Systems, and Manufacturing. 2013;7(3):362-76. doi: 10.1299/jamdsm.7.362.
Nanya T， Yoshihara H， Maekawa T.通过体素集成重建完整的 3D 模型。先进机械设计、系统和制造杂志。2013 年;7(3):362-76.doi：10.1299/jamdsm.7.362.

Zachos PK, Pappa M, Kalfas AI, Mansour G, Tsiafis I, Pilidis P, et al. Turbine Blading Performance Evaluation Using Geometry Scanning and Flowfield Prediction Tools. Journal of Power and Energy Systems. 2008;2(6):1345-58. doi: 10.1299/jpes.2.1345.
Zachos PK、Pappa M、Kalfas AI、Mansour G、Tsiafis I、Pilidis P 等人。使用几何扫描和流场预测工具进行涡轮叶片性能评估。电力与能源系统杂志。2008 年;2(6):1345-58.doi：10.1299/jpes.2.1345。

Segreto T, Bottillo A, Caggiano A, Martorelli M. Integration of reverse engineering and ultrasonic non-contact testing procedures for quality assessment of CFRP aeronautical components. Procedia CIRP. 2019;79:343-8. doi:10.1016/j.procir.2019.02.082.
Segreto T， Bottillo A， Caggiano A， Martorelli M. 整合逆向工程和超声波非接触式测试程序，用于CFRP 航空部件的质量评估。Procedia CIRP.2019;79：343-8。doi：10.1016/j.procir.2019.02.082.

Urata A, Date H, Kanai S, Gotoh T, Yasuda S. A Study on Rapid Reverse Engineering of Castings: A Separation of the Casting and Machining Surfaces from Scanned Triangular Meshes. Journal of the Japan Society for Precision Engineering. 2017;83(5):474-80. doi: 10.2493/jjspe.83.474.
UrataA，Date H，Kanai S，Gotoh T，Yasuda S.A 铸件快速逆向工程研究：铸件和加工表面与扫描三角形网格的分离。日本精密工程学会杂志。2017 年;83(5):474-80.doi：10.2493/jjspe.83.474.

Yang Y, Fang H, Fang Y, Shi S. Three-dimensional point cloud data subtle feature extraction algorithm for laser scanning measurement of large-scale irregular surface in reverse engineering. Measurement. 2020;151:107220. doi:10.1016/j.measurement.2019.107220.
YangY，Fang H，Fang Y，Shi S.逆向工程中大尺度不规则表面激光扫描测量的三维点云数据细微特征提取算法.测量。2020;151：107220。doi：10.1016/j.measurement.2019.107220.

Bonnici A, Akman A, Calleja G, Camilleri KP, Fehling P, Ferreira A, et al. Sketch-based interaction and modeling: where do we stand? Artificial Intelligence for Engineering Design, Analysis and Manufacturing. 2019;33(4):370-88. doi: 10.1017/S0890060419000349.
Bonnici A， Akman A， Calleja G， Camilleri KP， Fehling P， Ferreira A， etal. 基于草图的交互和建模：我们处于什么位置？用于工程设计、分析和制造的人工智能。2019 年;33(4):370-88.doi：10.1017/S0890060419000349。

Tanaka M, Terano M, Higashino C, Asano T. A Learning Method for Reconstructing 3D Models from Sketches, Computer-Aided Design & Applications, 2019;16(6):1158-70. doi: 10.14733/cadaps.2019.1158- 1170.
田中 M， 寺野 M， 东野 C， 浅野 T.A Learning Method forReconstructing 3D Models from Sketches， 计算机辅助设计与应用， 2019;16(6):1158-70.doi： 10.14733/cadaps.2019.1158-1170.

Montlahuc J, Shah GA, Polette A, Pernot JP. As-scanned Point Clouds Generation for Virtual Reverse Engineering of CAD Assembly Models, Computer-Aided Design & Applications, 2019;16(6):1171-82. doi: 10.14733/cadaps.2019.1171-1182
Montlahuc J、Shah GA、Polette A、Pernot JP。用于CAD装配模型的虚拟逆向工程的扫描点云生成，计算机辅助设计与应用，2019年;16(6):1171-82.doi：10.14733/cadaps.2019.1171-1182

Tashi, Ullah AS, Watanabe M, Kubo A. Analytical Point-Cloud Based Geometric Modeling for Additive Manufacturing and Its Application to Cultural Heritage Preservation. Applied Sciences. 2018;8(5):656. doi: 10.3390/app8050656.
Tashi， Ullah AS， Watanabe M， Kubo A. 基于分析点云的增材制造几何建模及其在文化遗产保护中的应用。应用科学。2018;8(5):656.doi：10.3390/app8050656.

Tashi, Ullah AS. Symmetrical Patterns of Ainu Heritage and Their Virtual and Physical Prototyping. Symmetry. 2019;11(8):985. doi: 10.3390/sym11080985.
扎西，乌拉AS.阿伊努遗产的对称图案及其虚拟和物理原型。对称。2019;11(8):985.doi：10.3390/sym11080985.

Tashi, Ullah AS, Kubo A. Geometric Modeling and 3D Printing Using Recursively Generated Point Cloud. Mathematical and Computational Applications. 2019;24(3):83. doi: 10.3390/mca24030083.
Tashi， Ullah AS， Kubo A. 使用递归生成的点云进行几何建模和 3D 打印。数学与计算应用。2019 年;24(3):83.doi：10.3390/mca24030083。

Ullah AMMS, Omori R, Nagara Y, Kubo A, Tamaki J. Toward Error-free Manufacturing of Fractals. Procedia CIRP. 2013;12:43-8. doi: 10.1016/j.procir.2013.09.009.
UllahAMMS，Omori R，Nagara Y，Kubo A，Tamaki J.迈向分形的无差错制造。Procedia CIRP.2013;12:43-8.doi：10.1016/j.procir.2013.09.009.

Tashi, Ullah AMMS, Kubo A. Developing a Human-Cognition-Based Reverse Engineering Approach. Proceedings of the JSME 2020 Conference on Leading Edge Manufacturing/Materials and Processing (LEMP2020), June 22-26, 2020, Cincinnati, OH, USA.
Tashi， Ullah AMMS， Kubo A. 开发一种基于人类认知的逆向工程方法。JSME 2020前沿制造/材料和加工会议论文集 （LEMP2020），2020 年 6 月 22 日至 26 日，美国俄亥俄州辛辛那提。