A flexible assembly retrieval approach for model reuse
Highlights
► A new assembly retrieval approach, in which users can use flexible queries to retrieve the assemblies from the product library efficiently. ► A multilevel assembly descriptor supporting various searching requirements which collects different levels of information. ► A matching method with well-balanced efficiency and discriminability. ► An indexing mechanism for accelerating assembly retrieval, especially the partial retrieval.
Introduction
In the information age today, the amount and value of information is increasing rapidly due to the widespread use of information technology. Specifically, as the computer-aided design (CAD) systems enjoy ever greater popularity in modern industries, a vast number of 3D digital models are generated and stored in the internet or enterprise repositories. These models always contain plenty of embedded knowledge worthy of utilization. Therefore, it will be a huge waste if the models, together with the embedded knowledge, could not be discovered and exploited to help practical works. Model retrieval technology, capable of searching out the models similar or relevant to a user input query from a huge library, is apparently conductive to the reuse of the abundant models and knowledge available worldwide.
In fact, model retrieval can be a dominant technology bringing about remarkable changes in the world. Various daily works can benefit from the model retrieval technology, such as design, study, exhibition and consultation. For example, designers always need to find models reusable for their design tasks at hand. The functional decomposition, geometrical implementation or even manufacturing cost estimation of a new design can all be accelerated by studying or reusing the design intents from similar models. Sometimes, when designers have just a rough idea in their minds, they can even provide a rough query model and use model retrieval to find some relevant models to inspire their designs. Since design is crucial to the success of a product in the market, modern industries can achieve huge gains with the increase of design efficiency granted by model retrieval. As Llewelyn [1] point out, the design work takes up only about 15%–20% of the total cost, but the quality of the design can decide 70%–80% of the total cost. Meanwhile, in statistics, model and knowledge reuse takes a considerable part in design activities. According to Gunn [2], only 20% of parts require completely new designs, while 40% of them are obtained by direct reuse and the other 40% through a modification of the existing designs. Therefore, it is urgent for modern enterprises and companies to possess model retrieval technologies which fulfill designers’ various requirements. Well developed and utilized model retrieval systems can enormously help enterprises and companies shorten the product lifecycles, which ultimately determines their successes in the global market.
Given that manual browsing is tedious and error-prone, there have emerged several kinds of tools for search so far, e.g. text-based search. However, as mentioned in [3], the text-based search has its limitations and is not always a good way to search engineering models. After the text-based search, methods for searching models based on contents have drawn considerable attention, and lead to numerous corresponding studies in the past decades [4], [5]. Specifically, in the engineering domain, shape-based 3D model retrieval has been well explored [6].
The shape-based methods mainly focus on analyzing the visual appearances of 3D models and matching the extracted shape descriptors between them. Although this paradigm may be effective in some part searching works, it cannot be directly used in the assembly retrieval. With respect to the assembly retrieval, which is the focus of this paper, it searches and reuses complex mechanical assembly models consisting of many sub-level parts or components. In the assembly retrieval, it is the product structure and the relationships between components that are really important, rather than the overall shapes of assembly models. In other words, two assemblies may be very relevant when their overall visual appearances are quite different. Therefore, the current shape-based methods are unable to tackle this problem well. This limitation of shape-based methods can also cause problems in some specific part searchings [7].
The search method in [3] begins to address the assembly retrieval in a way other than the shape-based methods. Besides the overall assembly statistics (e.g. the number of parts), the mutual relationships and structures existing in assemblies are also considered in this method. However, this method is yet to satisfy practical needs due to the following problems:
- (a)
The hierarchy in product structure is not considered. In high-level design such as top-down design or innovative design, the rough query with vague components and main relationships needs to be supported by an assembly retrieval system. This cannot be achieved without the hierarchical information.
- (b)
The semantics of assembly interfaces are not explored. The assembly interfaces between components in an assembly can exhibit many different forms under the same kinematical essentials. Many similar assemblies may be missed in the search if this information is not utilized.
- (c)
The indexing mechanism is absent. The efficiency will be a problem in case of an extremely large library of assemblies.
In this paper, a new assembly retrieval approach is presented with a view to overcoming the above-mentioned problems. In particular, a multilevel assembly descriptor capturing the main characteristics of assembly models is designed first. In this descriptor, the hierarchical assembly structure and the assembling semantics are extracted and maintained. Then graph-matching based algorithms are used to compare the multilevel assembly descriptors and calculate the corresponding similarities between assembly models. Finally, an indexing mechanism which supports sub-graph matching is given to accelerate the assembly retrieval. In the approach, the information arrangement and corresponding calculation during search is seamlessly integrated into a unified framework. Moreover, in the process of calculation, it is flexible to utilize different portions of the information for different application requirements. As a result, flexible queries can be well supported by our assembly retrieval approach. For example, while a conceptual designer inputs a rough query with abstract components and the kinematic-pair relationships defined between them to search out the assembly models similar in high-level concept, another detailed designer can input a more concrete query with additional information like shapes, layouts or geometric-matings defined. Sometimes the different levels of information can even be mixed together as users want or are accustomed to. And for sure, queries for general retrieval and partial retrieval (i.e. the query model can be a portion of the target model in the assembly structure) are all supported.
The rest of the paper is organized as follows. In the second section, we give a brief review of related works about CAD assembly retrieval. In Section 3 an overview of the flexible assembly retrieval approach is provided. Section 4 defines the multilevel assembly descriptor at length and illustrates the corresponding reasons and thoughts behind it. In Section 5 the details of matching and similarity assessment algorithms are given, while Section 6 describes the indexing and filtering mechanisms. Section 7 introduces the implementation details of the prototype system and analyzes the experimental results. Finally, we conclude the paper and present further work directions.
Section snippets
Related works
There are quite a few works dedicated to the content-based model retrieval. The works here we survey are those most closely related to mechanical engineering domain.
As is mentioned above, model retrieval works in engineering can be grouped into three main sub-aspects: the sketch retrieval, the part retrieval and the assembly retrieval.
The sketch retrieval methods mainly process digital images or vector drawings. Surveys [8], [9] analyze many relevant works in this area, and some latest typical
Overview of the flexible assembly retrieval approach
Fig. 1 shows the overview of our flexible assembly retrieval approach. It could be seen that our approach contains three main parts, i.e. the online processing, the offline processing and the assembly database. Here we give a brief description of each part respectively:
- a.
Assembly database
A multilevel assembly descriptor
In order to effectively achieve flexible assembly retrieval, a reasonable and comprehensive assembly descriptor involving various levels of data from high-level information such as kinematical properties to low-level information like geometric shape is a prerequisite.
In the work, we present a multilevel assembly descriptor for both queries and database models involved in assembly retrieval. The information in the presented assembly descriptor is multilevel in three aspects. First, it includes
Assembly matching
During the assembly retrieval, the query descriptor needs to be compared with the assembly descriptors in the database to get the similarities between them. Since the multilevel assembly descriptor contains abundant information, the matching process is divided into the following two main steps.
Assembly indexing and filtering
Although the VF2 sub-graph isomorphism algorithm used in the hierarchical graph matching is a quite efficient algorithm among the exact graph matching algorithms, it may still make the response time unacceptable when the amount of assemblies in the library becomes extremely large during practical assembly retrieval. Therefore, an efficient indexing structure for quick filtering of unmatchable models should be established for the assembly library to accelerate the assembly retrieval, especially
System
The proposed assembly retrieval approach has been implemented in a multi-module prototype system. The UI module (Fig. 19) is developed by using Microsoft Visual C# 2008, while the core module for the matching of descriptors (based on the VFLIB [60]), similarity assessment, indexing and filtering is developed by using Microsoft Visual C++ 2008, which is built as a win32 library invoked by the UI module during retrieval. Besides the two main modules, a C++/CLR module is developed as the
Conclusion and future works
In this paper, a flexible and effective approach is presented for searching assemblies in the product library. The multilevel assembly descriptor gathers different levels of information important for distinguishing assemblies. The hierarchical assembly structure and the semantic assembly interface can well preserve the implicit high-level design semantics hidden in assembly models. Moreover, the layout information and the shape information are also kept for better discriminability. Based on the
Acknowledgment
The authors are very grateful to the financial support from the National Science Foundation of China (No. 61173125).
References (62)
Review of CAD/CAM
Computer-Aided Design
(1989)- et al.
Content-based assembly search: a step towards assembly reuse
Computer-Aided Design
(2008) - et al.
Three-dimensional shape searching: state-of-the-art review and future trends
Computer-Aided Design
(2005) - et al.
Image retrieval: current techniques, promising directions, and open issues
Journal of Visual Communication and Image Representation
(1999) - et al.
A new approach to similarity retrieval of 2-D graphic objects based on dominant shapes
Pattern Recognition Letters
(1999) - et al.
Towards content-based retrieval of technical drawings through high-dimensional indexing
Computers & Graphics
(2003) - et al.
On visual similarity based 2D drawing retrieval
Computer-Aided Design
(2006) - et al.
Classifier combination for sketch-based 3D part retrieval
Computers & Graphics
(2007) - et al.
Structure-oriented contour representation and matching for engineering shapes
Computer-Aided Design
(2008) - et al.
Machining feature-based similarity assessment algorithms for prismatic machined parts
Computer-Aided Design
(2006)
Multiresolutional similarity assessment and retrieval of solid models based on DBMS
Computer-Aided Design
3D model retrieval using principal plane analysis and dynamic programming
Pattern Recognition
Case-based conceptual design information server for concurrent engineering
Computer-Aided Design
Case-based design for assembly
Computer-Aided Design
Managing digital libraries for computer-aided design
Computer Aided Design
A hierarchical data structure for representing assemblies: part 1
Computer-Aided Design
Storage and retrieval of description of mechanisms and mechanical devices according to kinematic type
Journal of Mechanisms
Automated conceptual design of mechanisms
Mechanism and Machine Theory
The mechanization of design and manufacturing
Scientific American
Content-based image retrieval at the end of the early years
IEEE Transactions on Pattern Analysis and Machine Intelligence
A survey of content based 3D shape retrieval methods
Multimedia Tools and Applications
3D solid model retrieval for engineering reuse based on local feature correspondence
The International Journal of Advanced Manufacturing Technology
Content-based multimedia information access
Sketch-based retrieval of complex drawings using hierarchical topology and geometry
Computer-Aided Design
A survey of shape similarity assessment algorithms for product design and manufacturing applications
Journal of Computing and Information Science in Engineering
Content-based retrieval of 3D models
ACM Transactions on Multimedia Computing, Communications, and Applications
Toward effective mechanical design reuse: CAD model retrieval based on general and partial shapes
Journal of Mechanical Design
Design reuse oriented partial retrieval of CAD models
Computer-Aided Design
Cited by (111)
Structure correspondence searching of CAD model using local feature-based description and indexing
2024, Pattern RecognitionInterface-Based Search and Automatic Reassembly of CAD Models for Database Expansion and Model Reuse
2024, CAD Computer Aided DesignBridging the Gap between Geometry and User Intent: Retrieval of CAD Models via Regions of Interest
2023, CAD Computer Aided DesignFuS-GCN: Efficient B-rep based graph convolutional networks for 3D-CAD model classification and retrieval
2023, Advanced Engineering InformaticsAn inexact subgraph matching algorithm for subpart retrieval in NC process reuse
2023, Journal of Manufacturing SystemsApplication of tensor factorisation for CAE model preparation from CAD assembly models
2022, CAD Computer Aided Design