Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
In recent years, the interdisciplinary topic of engineering change (EC) and its management (ECM) has gained increasing popularity within systems engineering benefiting also from the rise of attention towards concepts such as concurrent engineering, simultaneous design, product platform development, mass customization, and configuration management. The first widely acknowledged survey on ECM covering literature published between 1980 and 1995 identified only 23 articles [Wright, 1997]. Since then, the number of published works relevant for the field of ECM has risen continuously and significantly.
This paper contributes to the ECM research in three ways. First, it presents a holistic literature categorization framework which helps to structure research in ECM and related topics along the EC process. Second, it provides a relatively complete picture of research in ECM by positioning 384 journal articles and conference papers and 43 books, book sections, and reports in the proposed framework. Third, it presents a citation analysis for selected publications of the core ECM categories. The paper is structured in five remaining sections. Section 2 provides the background for the proposed categorization framework by defining EC and ECM and elaborating existing literature reviews and categorization frameworks. Section 3 introduces the proposed categorization framework. Section 4 presents the literature survey, positioning, and citation analysis; Section 5 discusses those results; and Section 6 concludes the paper.
2. BACKGROUND OF THE PROPOSED LITERATURE CATEGORIZATION FRAMEWORK
2.1. Relevance of EC in the Context of Engineering Design
The process of designing rarely starts from scratch, but rather by modification of existing products [Bucciarelli, 1994; McMahon, 1994; Cross, 2000]. On the basis of research effort, knowledge and skill, and creativity required in the designing process, a broad classification of design products into two types is without controversy [Pahl et al., 2007].
The first case where a new product is entirely designed from scratch is generally referred to as original design [Otto and Wood, 2001; Pahl et al., 2007]. Other terms used to address this type are novel design [Prebil et al., 1995] or creative design [Gero, 1990]. Original designs usually require much research work, knowledge and skill, and creativity [Pahl et al., 2007].
The second case where a product is designed by modification of an existing one is known as evolutionary design [Frazer et al., 2002; Kicinger et al., 2005]. This design type is initiated and driven by ECs. Some authors distinguish between two different types of evolutionary design, e.g. routine design and innovative design [Gero, 1990], or alternatively, variant design and adaptive design [Otto and Wood, 2001]. Variant design refers to designs with different values of specific parameters of the design elements, whereas adaptive design refers to designs with different specific design elements [Otto and Wood, 2001]. Evolutionary designs usually require less effort than original design.
It is generally accepted that the vast majority of designs are adaptive and of variant design and thus per definition initiated by ECs. However, original design projects are also usually subject to ECs during the product development (PD) phase and along their life cycle as a means to rectify errors or meet changing requirements. Hence, “it is absolutely necessary to understand changes and to have a good grip on them, as the entire PD process can be described as a continuous change management process” [Fricke et al., 2000: 177].
“An engineering change (EC) is a modification to a component of a product, after that product has entered production” [Wright, 1997: 33].
“[Engineering changes are] the changes and modifications in forms, fits, materials, dimensions, functions, etc. of a product or a component” [Huang and Mak, 1999: 21].
“Engineering change orders (ECOs) [are] changes to parts, drawings or software that have already been released” [Terwiesch and Loch, 1999: 160].
“Engineering changes are changes and/or modification in fits, functions, materials, dimensions, etc. of a product and constituent components after the design is released” [Huang et al. 2003: 481].
“An engineering change is an alteration made to parts, drawings or software that have already been released during the product design process. The change can be of any size or type; the change can involve any number of people and take any length of time” [Jarratt et al., 2004c: 268].
The difference between these definitions can be visually demonstrated by mapping their coverage against a product lifecycle as proposed by Pahl et al.  and Ulrich and Eppinger  (see Fig. 1). This mapping and comparison with an “ideal definition” shows that none of these definitions is comprehensive and distinctive enough—“comprehensive” referring to both the scope of the definition (represented by the height of the grey filling of the bars) and the lifecycle coverage (represented by the length of the bars) and “distinctive” referring to the distinguishing of ECs from design iterations (represented by the small bars with gaps in between).
As depicted in Figure 1, Wright's definition covers only changes during the production stage and ignores changes during the PD and testing stages. Huang and Mak  include in their definition a wider scope of ECs, but leave out the time aspect required to distinguish ECs from design iterations. In contrast to ECs, the latter usually occur before the release of documents. Iterations are inevitable for creative design processes which are characterized by a concurrent exploration of problem and solution space [Lawson, 1980; Dorst and Cross, 2001]. Wynn et al.  explored the nature of iterations and categorized them into rework-related, convergence-related, refinement-related, and repetition-related. The definition given by Terwiesch and Loch  draws the line between ECs and design iterations by restricting the time of ECs to the post-release phase. In addition, it enhances the scope of ECs to embedded software, which is essential for most modern complex products. Also Huang et al.  add the time aspect to their initial definition. Jarratt et al. [2004c] give the most comprehensive definition, but omit functions from the scope.
In order to cover a wide range of research on ECM, a broader definition of ECs which draws on the definitions by Huang et al.  and Jarratt et al. [2004c] is used in this work:
ECs are changes and/or modifications to released structure (fits, forms anddimensions, surfaces, materials etc.), behavior (stability, strength, corrosion etc.),function (speed, performance, efficiency, etc.), or the relations between functions andbehavior (design principles), or behavior and structure (physical laws) of a technicalartefact.
The adjective technical in this definition is used in the broader sense to differentiate ECs from changes to non-technical artefacts such as social (e.g. laws), artistic (e.g. painting), or architectural (e.g. building construction). Artefact is an umbrella term which may refer to a single part, a component, an assembly, a system, or a whole product. Software and controller units may be parts of such technical artefacts and are included in this definition. Changes to the manufacturing process or tools are not automatically ECs but can lead to those when they entail changes to released product attributes.
This definition is in consensus with the five modes of incremental change in design as developed by McMahon  who distinguishes between “explicit” and “implicit attributes.” Explicit attributes are those required on the drawings, technical documents etc. in order to produce the product, whereas implicit attributes emerge from them. The former determine the product's structure and the latter its behavior and function. Based on a framework consisting of these attributes, a design space, and product requirements, McMahon  defines five modes of incremental change in design which lead to development of designs over time. These modes can be mapped to the product domains—structure, behavior, and function—as shown in Table I.
Table I. Mapping of McMahon's Modes of Design Change to the Product Domains
ECM refers to the organization, control, and execution of ECs [Jarratt et al., 2011] and covers the product life cycle from the selection of a concept to the wind-down of production and support. The goals of ECM are to avoid or reduce the number of engineering change requests (ECRs) before they occur, to select their implementation effectively when they occur, to implement required ECs efficiently, and to learn from implemented ECs. These five goals are termed Less, Earlier, More Effective, More Efficient, and Better by Fricke et al. . ECM can be regarded as the core of the larger configuration management [Jarratt et al., 2011]. The latter deals with establishing and maintaining consistency of a product's performance, functional, and physical attributes with its requirements, design, and operational information throughout its life [ANSI/EIA-649, 1998]. Configuration management is an integral discipline within systems engineering [see, e.g., Shishko and Chamberlain, 1995; Sage and Rouse, 1999]. However, the focus of this work is ECM.
2.3. EC Process
EC processes have been proposed at different levels of abstraction. Dale  split the process into two stages: procedure to approval of ECs and procedure onapproval. Maull et al.  proposed a five-step IDEF process model including the steps (1) filter proposal, (2) design investigation, (3) appraise design, (4) authorizechange, and (5) execute change. Rivière et al.  proposed three stages, ECproposal, EC investigation, and EC embodiment, and more detailed process steps for each stage.
A four-stage model for the formal process of ECs was proposed by Lee et al. , including the stages (1) initiating an ECR, (2) evaluating the ECR, (3) issuing engineering change orders (ECOs) to relevant participants, and (4) storing and analyzing the ECOs for management purposes. Jarratt et al. [2004c] proposed a more comprehensive three-stage process including six process steps. This generic process covers the complete life cycle of ECs, from their initiation through their implementation to their review. It is structured in six steps and three stages as shown in Figure 2.
The first four process steps are similar to the first four process steps from Rivière et al. ; the fifth process step covers all three process steps of the third stage of the latter; and the last process step includes an important learning step of modern EC practice [Lee et al., 2006], which is missing in the process of Rivière and colleagues. Furthermore, two of the most likely iterations and four possible break points, at which the change process can be brought to a halt by the control mechanism, are marked in the process chart.
2.4. Existing Literature Reviews and Categorization Frameworks for ECM
The first widely acknowledged survey on ECM, which covered literature published between 1980 and 1995, was conducted by Wright , who identified only 15 core papers and eight further articles, after studying the citations of those core papers. Wright suggested a tree-structured ECM research categorization framework which differentiates between two main categories of related research: (1) computer tools for the analysis of EC problems and synthesis of solutions and (2) methods for the reduction of EC impact. These categories were further subdivided as shown in the Figure 3. Another, similar tree-structured framework was proposed by Ouertani .
More recently, Jarratt et al.  have conducted a literature review which provides a broad and deep up-to-date coverage of the field including more than 100 relevant publications. The survey highlights key EC articles published until 2010 and aims to provide an up-to-date coverage of the relevant ECM topics. Their categorization framework classifies relevant publications into the three main categories, (1) Process, (2) Tool, and (3) Product, and two additional categories for (4) General studies, and (5) Strategies and methods to cope with ECs. In contrast to the tree-structured categorization frameworks suggested by Wright  and Ouertani , these categories overlap and allow an article to be assigned to multiple categories.
A systematic survey on literature published between 2005 and 2010 including statistical analysis and an overview of different research groups around the globe can be found in Ahmad  and Ahmad et al. . The survey considers 314 publications—mainly conference papers from the International Conference on Engineering Design (ICED), the International Design Conference (DESIGN), and the ASME Design Engineering Technical Conference (DETC)—for statistical analysis and lists the reference of the main publications. The four main categories proposed are (1) Change management, (2) Design taking into account of changes, (3) Organizational changes, and (4) Others. Furthermore, according to the addressed domains of design, the literature is categorized into (1) Requirements, (2) Function models, (3) Component/subsystem models, (4) Detailed design process, and (5) Cross domain models. Both category dimensions are used to build different clusters. However, neither a complete list of the analyzed publications is provided nor are the clusters clearly distinguished against each other.
Literature surveys looking at certain aspects of ECM are presented by Huang and Mak  and Rouibah and Caskey . Huang and Mak  reviewed computer aids for EC control. Rouibah and Caskey  reviewed ECM literature relevant for concurrent engineering and categorized them into (1) Survey research or field research, (2) Industrial case studies, (3) Methods and frameworks for implementation, and (4) Tools and IT solutions. There are other literature reviews on related topics, most noticeably on product development process models by Browning et al.  and Browning and Ramasesh , on product platforms by Simpson , and on concept selection methods by Okudan and Tauhid .
Other categorization schemes can be found in conference proceedings. For example, in ICED, since 2009 design research has been categorized into the nine categories: (1) Design Processes, (2) Design Theory and Research Methodology, (3) Design Organization and Management, (4) Product and Systems Design, (5) Design Methods and Tools, (6) Design for X and Design to X, (7) Design Information and Knowledge, (8) Human Behavior in Design, and (9) Design Education. Although ECM papers are more likely to be part of the groups 1, 3, 5, and 7, relevant papers may be found within all nine groups. A concurrent classification scheme for research in engineering design was proposed in a keynote presentation of ICED 2011 entitled “Design Research: Embracing the Diversity” by McMahon . This applied faceted classification using eight facets—including addressed time period, number of activities and issues, number of people, and product lifecycle phase—to classify the 340 conference papers from ICED 2009—excluding the Design Education papers. He found that the patterns differ between the eight categories. For the group (1) Design Process, where in ICED 2009, the majority of the ECM related papers were assigned to, the pattern suggested studies of weeks or a few months, medium range of number of activities and issues, primarily focused on single people and small teams, components and assemblies, high level of abstraction, medium level of originality, conceptual design phase, and using preferably modeling, experimental methods, and tools. Such a faceted classification of related ECM publications might deliver additional insights.
In summary, there are several literature reviews on ECM available, but none of them provide a comprehensive up-to-date coverage as intended in this paper. They either consider only a given period (e.g., the reviews by Wright  and Ahmad et al. ), or focus on selected articles (e.g., the review by Jarratt et al. ), or on particular aspects of ECM (e.g., the reviews by Huang and Mak  and Rouibah and Caskey ). This paper supplements those existing reviews with a nearly-complete coverage of all relevant publications in the field of ECM as shown in Figure 4 This schematic Euler diagram shows the overlap of the references of the three literature review based publications—Jarratt et al. , Wright , and Rouibah and Caskey —with the list of categorized publications in this paper. The diagram suggests that the categorized list of publications in this paper includes the majority of the references of each of those three publications plus 300 additional publications which are not cited by either one of them. The references included in the reviews but not categorized are either publications with low relevance for ECM or reference types which are not included in the categorization such as e.g. Ph.D. theses, technical standards, or homepages. The proposed categorization frameworks are either too narrow, focusing only on “core papers” (e.g., the framework by Wright , Ouertani , and Rouibah and Caskey ), not disjunctive enough (e.g., the framework by Jarratt et al. ), or too specific (e.g., the clusters by Ahmad et a. ). There is no framework which allows a disjunctive categorization and comprehensive coverage of related literature in ECM and related topics in its broad context. This is the aim of the framework presented in the next section. This framework complements the existing literature reviews with an illustration of the position of the publications in the context of the big picture of ECM.
3. A HOLISTIC CATEGORIZATION FRAMEWORK FOR LITERATURE ON ECM
3.1. Framework Overview
In pursuit of providing a broad overview of the state of research in ECM and allowing a more precise positioning of relevant publications, a new categorization framework has been developed (see Fig. 5). This framework is holistic as it covers not only ECM but also related cross-disciplinary areas. In its core, it includes the EC process from Jarratt et al. [2004c]. The purpose of this framework is to help reveal research gaps and opportunities by capturing and visualizing the current research status along the ECM process. The main blocks of the framework are (A) Prechange stage, (B) In-change stage, (C) Postchange stage, and (D) General studies & surveys. They are defined below.
3.2. (A) Prechange Stage
Research in the prechange stage is concerned with concepts to prevent or to ease the implementation of ECs before they occur. Referring to the five goals of ECM mentioned above (i.e., Less, Earlier, More effective, More efficient, Better) research in this area predominantly focuses on Less. This block is subdivided into the categories (A.1) People-oriented, (A.2) Process-oriented, and (A.3)Product-oriented research. People-oriented research deals with training and development of designers and other employees across the whole process chain in order to create awareness of ECs, and provide the people with knowledge and skills of how to handle them. Process-oriented research focuses mainly on PD process modeling and optimization. Product-oriented research focuses mainly on the product architecture and covers concepts such as design for variety, design for changeability, axiomatic design, robust design, set-based design, systems architecting, requirements management, and flexible product platforms.
3.3. (B) In-Change Stage
In-change research is concerned with methods, tools, systems, strategic guidelines, and organizations to handle ECs when they occur. Referring to the five goals above, research in this area predominantly focuses on Early, More effective, and Moreefficient. This block is subdivided into the categories (B.1) Organizationalissues, (B.2) Strategic guidelines, (B.3) ECM systems, (B.4) Methods & ITtools, and (B.5) ECM process. While the other categories address the whole ECM process, Methods & IT tools are more specific and targeted at certain process steps. The Methods & IT tools category is therefore further divided into subcategories along the generic ECM process as suggested by Jarratt et al. [2004c].
3.4. (C) Postchange Stage
Research in the postchange stage is concerned with the ex post facto exploration of effects of implemented ECs on (C.1) Delays, (C.2) Cost, (C.3) Quality, (C.4)Premanufacturing stage, (C.5) Manufacturing & postmanufacturing stage, and (C.6) General sources & impacts. To keep the categorization mutually exclusive, the last category includes all publications which address more impacts and could be categorized into two or more of the categories (C.1)–(C.5). Referring to the five goals above, research in this area predominantly aims at Better.
3.5. (D) General Studies & Surveys Related to ECM
General studies & surveys cover research which explores the discipline of ECM and related topics as well as general surveys about ECM practice in industry. They can be subdivided into the categories (D.1) People-, process-, product-oriented and (D.2) ECM-oriented.
3.6. How To Position Publications
The assignment of publications to the categories should be based on their main contribution to ECM. The framework categories are mutually exclusive and collectively exhaustive. Thus, each contribution can be assigned to exactly one category. Most journal articles and conference papers have one main contribution to ECM which facilitates an unambiguous assignment. However, some have multiple distinct contributions, allowing them to be assigned to multiple categories. This framework could also be applied to position other more comprehensive publications such as books and theses. As those kinds of publications usually address more diverse topics, they can be assigned to multiple categories.
4. LITERATURE SURVEY AND CATEGORIZATION
4.1. Scope of the Literature Survey
The aim of this categorization is to provide a comprehensive and up-to-date picture of research that has been undertaken in the field of ECM in its wide context. The respective list of publications should serve as a useful database for both researchers and managers in the field of ECM. To account for the originality and quality of publications, only journal and conference papers have been considered. However, to provide a full list of references, in a second separate categorization, books and book sections have been positioned. Since there are already several limited literature reviews on the core field of ECM, which includes the blocks (B), (C), and (D.2) in Figure 5, this review aims to provide a relatively complete picture of research in those blocks. The wider context of ECM, which includes the blocks (A) and (D.1) in Figure 5, is related to other research fields such as organization theory, project, process, and product management, process, and product design. Therefore, only the most relevant publications concerning ECM are covered in those blocks, and the list is not complete.
The collection of publications followed four phases. First, a systematic search was conducted for the period from January 1996(after the survey by Wright ) to September 2011. In this phase, the search included the following journals and conference proceedings:
Journals: Research in Engineering Design; Journal of Engineering Design; Design Studies; IEEE Transactions on Engineering Management; Product Innovation Management; Computers in Industry; Systems Engineering; Artificial Intelligence for Engineering Design, Analysis and Manufacturing; and International Journal of Design Engineering.
Conference proceedings: ICED; DESIGN; DETC; and International DSM Conference.
This phase started by longlisting publications which included the word “change” in their title or abstract and progressed by shortlisting those referring by change to EC. Second, the results of the systematic search were completed with the references of existing literature surveys. Third, the key publications were cross-referenced. Fourth, the list was completed with an open search for the words “engineering change” using IEEE Explore, SpringerLink, Scopus, and Google Scholar.
The final list of journal articles and conference papers selected for the categorization includes 384 publications from more than 110 different sources. To provide a more complete list, conference papers which were superseded by journal articles were kept in the list. A distribution of the number of publications over year by type is shown in Figure 6: 192 (50%) of the publications are journal articles and 192 (50%) conference papers.
4.3. Categorization Approach and Results
The positioning of the papers has been completed in three steps. In the first step, a rough categorization to the main blocks (A), (B), (C), or (D) of the framework was conducted based on titles and abstracts. In the second step, the contents of all papers were screened to create a more precise categorization. Finally, in the last step, publications with multiple distinct contributions were positioned according to their primary and secondary contributions. The results are represented in the following depictions. Table II provides the distribution of the number of publications over category and year.
Table II. Distribution of Number of Journal Articles and Conference Papers over Year by Category According to Their Primary Contribution
Figure 7 visualizes the respective publications within the framework, and Tables III and IV include the associative details. As discussed above, some publications have multiple distinct contributions and could be assigned to different categories. Therefore, the publications which address ECM directly, i.e. from the blocks (B), (C), and (D2) in Table II, were additionally categorized according to their secondary contribution if such a second contribution was distinguishable within the publication. Furthermore, to provide a more complete list of publications, in a second attempt books, book sections, and other reports (see Table IV for details) were categorized separately according to their primary contribution.
Table III. Details of Categorized Journal Articles and Conference Papers
Table IV. Details of Categorized Books, Book Sections, and Other Reports
4.4. Citation Analysis
To analyze the links between the publications, a citation analysis has been conducted. This analysis was limited to publications categorized within the core field of ECM, i.e., from the categories (B), (C), and (D2). The categories (A) and (D2), which are more related to other research fields and have a wider scope than covered by the publications listed above, were excluded from this analysis. As citation database for this analysis the bibliographic database Scopus was used. Scopus covers publications which go back as far as the early 19th century but includes lists of references only for publications from 1996 or later. As a result, citation records are counted only from 1996 onwards.
First, all journal articles and conference papers listed in the categories (B), (C), and (D2) were searched in Scopus. From the 247 publications (112 journal articles and 135 conference papers) listed in Table II, 144 publications (99 journal articles and 45 conference papers) were found in Scopus, i.e., 88% of the journal articles and 33% of the conference papers listed in Table II. As Scopus does not include papers from ICED and DESIGN, the majority of missed conference papers come from these biennial conferences. Publications number 191 [Hauser and Clausing, 1988] and 321 [Griffin, 1997] were found in Scopus but were excluded as outliers from this analysis because of their high number of received citations originating mostly from other research areas. An overview of the number of publications over year by type is presented in Figure 8. Second, for this list of 144 publications, a list of citations over year as well as an h-graph was generated using Scopus. The overview of the number of citations over year by type is presented in Figure 9. The h-graph is in Figure 10. The respective h-index is 20, i.e., of the 144 publications considered for the h-index, 20 have been cited at least 20 times. A discussion of the h-index can be found in Hirsch . Those top 20 publications can be found in Table V. Furthermore, the cumulated number of citations was divided by the cumulated number of publications to generate an average number of citations per publication. This metric shows a steady increase from 0.4 in 1996 over 5.5 in 2006 to 9.3 in September 2011 and has a linear sloped graph.
Table V. Top 20 Most Cited ECM Publications in Scopus as of 30 September 2011
Third, for each of the 144 publications a list of references was generated. The majority of these lists (119 publications) were extracted from Scopus. For publications where such a list was not available in Scopus, e.g., all publications of 1995 and earlier, the references were retrieved manually. These lists of references were used to track the “who cited whom” information. For this purpose, a code was written which used the title of publication as identification key and searched the links. The title of publication was used as identification key for two reasons. First, it appears uniformly in all common citation formats, and, second, it is most likely to be unique for each publication. As the majority of the reference lists were downloaded from Scopus with uniform format and consistent titles, the frequency of error is considered to be low. However, exemplary manual checks were conducted, and exceptions were rectified manually.
The results of this analysis are presented in a design structure matrix (DSM), which includes the 144 publications as its elements and maps the “who cited whom” information using binary values of 1 (link available) and 0 (no link). Reading along a row of a given publication shows which other publications this has cited, hence a part of its list of references. Reading along a column of a given publication shows in which other publications this has been cited. The publications were ordered according to their date of publication; therefore, apart from one exception where two publications were published within a short term by the same authors—Loch and Terwiesch—the lower diagonal values of the DSM are all 0. The literature review based publications from Jarratt et al. , Rouibah and Caskey , and Wright  are highlighted and most of the top 20 publications are indicated by the labels. For further analysis, the DSM was imported into the Cambridge Advanced Modeller (CAM), a software program for modeling and analysing dependencies and flows in complex systems (for more information about CAM, see online under http://www-edc.eng.cam.ac.uk/cam or [Wynn et al., 2010]). A screenshot of this DSM is depicted in Figure 11. Although its details are not readable due to the small resolution, the DSM provides a valuable and comprehensive overview of the interrelations of the 144 publications in Scopus. The software tool can also generate a force-based graph of this DSM supporting graphical analysis of the corresponding network.
Finally, in order to highlight the categories previously defined and the citation links between them, the DSM were clustered based on those categories. The corresponding collapsed category-category DSM with the sums of citations per category is presented in Table VI.
Table VI. Collapsed Category-Category DSM Showing the Citation Links between the 144 Publications in Scopus within 16 Categories
5.1. Distribution of the Number of Publications over Year
The distribution of the publication numbers in Figure 6 shows a discontinuous increase until 2007 with a peak of 50 publications in 2007 and a continuous decrease after 2007 with 28 publications in 2010 and 19 in the first three quarters of 2011. This reflects also the distribution in each of the two subgroups: journal articles and conference papers. From this distribution, it could be concluded that the interest in ECM research steadily increased until 2007, achieved its peak in 2007, and has been decreasing since then, but still remains at a higher level compared to the period before 2000. A deeper analysis of the numbers of publications by source shows that 20 out of the 50 publications in 2007 came from ICED 2007 and were mainly submitted by European researchers. This conference took place in Paris and might have attracted many ECM papers partly also due to traveling convenience for the major European research groups. Just as the number of published conference papers for a given year depends on the parameters of the conferences, the number of journal articles depends on journal parameters such as the number of reviewers, the backlog, special issues etc. Therefore, any conclusion from the distribution of the number of publications over time on the level of interest in ECM is difficult. However, the pattern suggests an overall increase of research interest in ECM which is without controversy. Wright's review which was published in 1997 possibly contributed to this positive development leading to a significant increase of journal papers in 1999.
The number of conference papers is concentrated in the period after 2000 with zero conference papers between 1985 and 1994. This is due to the fact that the majority of considered publications until 1995 stem from Wright's  review which does not include any conference papers between 1985 and 1994. Although Wright's list of publications was completed by cross referencing of key publications and by the open search, no conference papers published in that period were found. Thus, it can be assumed that little work on ECM was published in conferences before 1995 and that ECM did not become established as a perennial topic within relevant conferences on engineering design until the early 2000s.
5.2. Distribution of the Number of Publications over Categories and Year
The overview of the categorization results in Table I shows an irregular distribution of the total number of publication over categories. This can be traced back to mainly two reasons. First, the scope of the categories is different. For example, the categories (A.1), (A.2), (A.3), and (D.1), which cover ECM related research from other specific research areas, are wide-open, while the borders of the other categories are more stringent. Second, the research challenges within the categories are different. For example, while change classification is straightforward, change prediction and impact analysis still pose a huge challenge for research. Authorization, for instance, is an important step but could be considered more as a gate at the end of the process step Impact analysis than a separate process step. The authorization decision is based on the results of the impact analysis. Therefore, publications addressing authorization primarily focus on impact analysis. There is no publication which in the first place addresses authorization issues, such as, what is the authorization procedure, who is involved in decision making, what procedures are required to authorize a change, when is which type of authorization required, how do tools support authorization, what information is required for decision-making. As a result, the category B.4.4 in Figure 7 is empty; Harhalakis , and Rouibah and Caskey  contribute to this category as a secondary focus as shown in Figure 7. However, to remain consistent with the well-accepted ECM process proposed by Jarratt et al. [2004c], all six process steps are considered as separate categories within the framework.
Considering both of these factors, the profile discloses the following insights:
Research in (A) Prechange stage mainly focuses on (A.3)Product-oriented and (A.2) Process-oriented measures to reduce the number and impact of ECs and less on (A.1) People-oriented measures.
Research in (B) In-change stage mainly focuses on (B.4.3) Impactanalysis, (B.4.6) Documentation and review, and (B.3) ECM systems, while (B.1) Organizational issues and (B.4.5) Implementation of ECs are infrequently addressed.
Research in (C) Postchange stage mainly focuses on (C.6) Generalsources and impacts, impacts on (C.4) Premanufacturing stage, and (C.1)Delays. There is a lack of research on the impacts of ECs on (C.3) Quality and (C.5) Manufacturing & postmanufacturing stage.
The distribution of number of journal articles and conference papers over year by category according to their primary contribution in Table II shows different patterns for the categories. While (B.4.6) Documentation & review, (B.5) EC process, and most of the postchange analysis categories (C.1–C.6) have received continuous attention during the period from the early 1980s to the present, (B.3) ECM systems, (B.4.3) Impact analysis, and (B.4.5) Implementation were not addressed as late as the early 2000s.
The Scopus database includes more than half (144 of 247) of all publications categorized within the core ECM field. Since 88% (99 of 112) of the listed journal articles and only 33% (45 of 135) of the listed conference papers are covered by the database, the citation analysis conducted above can be regarded as representative only for the journal articles but not for the conference papers. However, as journal articles have in general a higher impact, the results of this analysis are justifiable and implications for ECM are reasonable.
The distribution of the numbers of publications in Scopus over time in Figure 8 suggests overall an increase, considering that 2011 is incomplete and does not include any conference papers. The average number of publications is 4.0 in the 5-year period 1996–2000 and 11.4 in 2006–2010. Other than the numbers in Figure 6, there is no peak in 2007, which supports the discussion in Section 5.1.
The distribution of the number of citations in Scopus over year in Figure 9 suggests increasing number of citations since 1996 with an exceptional drop in 2007. This drop can be explained by different factors such as the number of ECM publications, the number of references of those publications, and the number of citations from publications outside of the core ECM field. The number of publications considered drops from 12 in 2006 to 9 in 2007 and leaps back to 12 in 2008 (see Fig. 8). The publications considered for the analysis have an average number of references of 33.4 in 2006, 20.4 in 2007, and in 21.7 in 2008. The portion of the number of citations coming from the publications of the core ECM field is on average 33% (446/1,337). Therefore, the drop in 2007 is also linked to a drop in the number of citations coming from publications outside the core ECM field.
The average number of citations per publication in Scopus suggests a linear increase. This linear trend is expected because the cumulated numbers of both the citations and the publications increase progressively. As of writing, the distribution of the number of citations per publication is right skewed with an average of 9.3 and a (minimum; median; maximum) of (0; 3.0; 72). The gradient of the linear graph suggests on average 0.58 citations per publication per year, that is, one citation per publication every 1.72 years. The h-graph in Figure 10 shows the distribution of the number of citations in Scopus over publications (i.e., documents) and indicates only 20 publications with 20 or more citations (i.e., h-index 20). As there are some ECM-related researchers with an h-index of 20 or more in Scopus (e.g., L.P. Zhang: 36, L. Chen: 23, G.Q. Huang: 20. C. Terwiesch: 20), this is a relatively low h-index for a research field. These findings underline that ECM is a relatively young and yet low-cited research field.
Table V shows that the top 20 ECM publications are well distributed over different categories and researchers. However, the category ECM systems and the researcher Huang stand out due to multiple occurrences.
Figure 11 suggests a well-distributed number of citations between the selected publications available in Scopus over the whole period, with some more populated and a few empty rows and columns. The most populated columns identify the most cited publications, while empty columns show publication with no citations. The most populated rows identify the publications with the most references, while empty rows show publication with no references.
Overall, this distribution suggests that most ECM publications refer to other, existing publications and that the number of citations per publication increases with the time, i.e., newer publications cite in average more references than older publications. From this, it can be concluded that the ECM research field builds on existing knowledge from the past. Furthermore, the distribution shows that there are some key publications which receive lots of citations (most populated columns) and that there are some literature review based publications which use a lot of references (most populated rows). Despite the area-wide distribution of links in the upper-diagonal matrix half, on the top right corner of the DSM, a section with many empty cells is located. A more detailed analysis shows that this is related to the literature review by Wright , which seems to be cited as replacement for its references; 15 times (out of 26 citations) it appears as a citation without any citations of its references. This is significantly lower for other literature review based publications such as Rouibah and Caskey .
The total number of actual links between publications (Fig. 11) adds up to 446. Jarratt et al.  have with 41 the maximum actual number of references (sum of rows), while Rouibah and Caskey  have 13, Wright  has 15, and the overall average is 3.1 actual references from the listed publications in Scopus. Dividing these actual numbers of references by the possible numbers of references for each publication shows a more detailed picture of reference coverage per publications. This reference coverage ratio has a lower variation than the actual numbers of references and has rather remained stable over the period. Wright  has the maximum coverage with 57.7%, while Rouibah and Caskey  have 23.6% and Jarratt et al.  29.9%. The average reference coverage is 4.3%, suggesting that on average 4.3% of available ECM publications in Scopus are cited by each new ECM publication in Scopus; for 2011, this corresponds to 6.0 references per publication. Assuming an average number of 30 references for a given publication and taking into account that only 58.3% (144/247) of the core ECM publications are listed in Scopus, this suggests that one third of its references come from the core ECM field. This quantitative finding supports the conclusion drawn above, that the ECM research field builds on existing knowledge from the past.
The cluster analysis in Table VI shows for most categories a higher number of intracategory citations (diagonal values) than the single intercategory citations (off-diagonal values). Although, the statistical significance of this distribution is not proven, it verifies the categorization. Furthermore, as the categories (B.3)EC systems, (B.4.4) Implementation, and (D.2) ECM-oriented studies &surveys cite almost all other categories, the DSM in Table VI cannot be brought into a triangular form. This means that the categories are interlinked, and there is no sequential order of the categories in terms of citations (i.e., input/ output).
5.4. Research Gaps and Opportunities
From the above gained insights, the following research opportunities can be drawn:
1. People-oriented measures to avoid ECs. Having studied the relation of designer's manufacturing knowledge and the occurrence of ECs, Saeed et al.  concluded that organizations can reduce the number of ECs when they train designers in manufacturing-knowledge. Other studies discuss the role of communication in collaborations, e.g., between embodiment design and simulation [Maier et al., 2009], within trust networks [Atkinson et al., 2011], and between design teams [Lindemann et al., 1998; Eckert et al., 2000]. However, there is a need for further research on how ECs can be avoided through people-oriented measures. Such measures include optimization of organization and team structures in collaborations, improvement of both the quality and frequency of communication and knowledge sharing among designers, between designers and other disciplines and with customers and stakeholders, development of designers' technical know-how and soft skills, work life balance and working conditions, and quality control measures.
2. Organizational issues. DiPrima  discussed the role of an EC coordinator and the formation of an EC board. Other researchers reported on ECM organizations of their case study companies, such as Balcerak and Dale , Pikosz and Malmqvist , Terwiesch and Loch , Eckert et al. , and Tavcar and Duhovnik . Furthermore, organizational issues of product development are picked up in books, such as Leech and Turner  and Clark and Fujimoto . However, there is a lack of research on organizational issues of ECM. Such research could investigate which types of ECM organizations are currently applied in practice, evaluate and compare their efficiency against each other, compare them with organizational structures of other disciplines, and come up with guidelines of how to improve ECM organizations. Furthermore, the dependency of ECM organization on different influencing factors such as the product, the number and size of involved teams, and the specifics of the business and market environment could be investigated.
3. Implementation of ECs. A few authors have addressed the implementation of ECs. Bhuiyan et al.  compared two implementation policies—individual implementation of changes as they occur or implementation in a batch—and found that batching of ECs was superior. Ahmad et al. [2010b] found as a result of a simulation that there is an optimal batching size of ECs with minimum impact on the average project delay. Barzizza et al.  proposed a categorization of ECs according to their urgency of implementation in three groups and suggested optimal implementation times for each group. Other simulations investigating the complex interrelations between EC tasks and designers in a collaborative network to support effective and efficient ECM processes include Gärtner et al. , Li and Moon , Eckert et al. , Wynn et al. , and Reddi and Moon [2011a]. Some authors have also discussed workflow models [e.g., Chinn and Madey, 2000; Liu et al., 2004; Qiu and Wong, 2007] and strategies to overlap coupled activities and reorganize the execution of design task [e.g., Ouertani, 2008] in the context of ECM. Some approaches on dealing with iterations [e.g., Krehmer et al., 2008; Langer et al., 2011] as well as much of the work on process management listed under the category A.2 can also be applied to support the implementation of ECs. However, there is still a need for further research to help planning and implementing EC tasks which compete with other running projects for resources. Such work could elaborate EC tasks sequencing and routing, assignment of resources, management of timing and deadlines, and scheduling of all EC projects and running projects.
4. Impacts of ECs on quality. Hua and Wemmerlöv  explored the impact of a company's product change frequency on the product performance, i.e., sales and market share, based on a survey with 55 U.S. companies in the personal computer industry and found a positive correlation between them. Using iteration as a measure of design quality, Van Wie and Flechsig  proposed a method which incorporates the principles of verification and change management to address the quality control of engineering artefacts in early stage design. Yet there is a need for research to investigate the impact of ECs and ECM on the actual quality of products and processes as well as on the perceived quality in terms of customer satisfaction and brand image.
5. Impacts of ECs on Manufacturing & postmanufacturing stage. The impacts of ECs on manufacturing raised the attention of ECM researchers from the early years of ECM research [e.g., Heumann, 1983; Williams, 1983]. However, since the work by Coughlan  there has not been any publication focusing primarily on the analysis of impacts of ECs on manufacturing and postmanufacturing stages. Most authors refer to the “Rule of Ten” or similar studies to underline the impact of ECs on the manufacturing and postmanufacturing stages [e.g., Clark and Fujimoto, 1991; Fricke et al., 2000]. However, these studies are focused on cost or delays but do not discuss other effects of ECs such as variations in production program and resource planning, changes in tooling, stock keeping, logistics, and services. As companies increasingly offer their products under full service agreements which require them to care for service and maintenance throughout the product life-cycle, the impact of ECs on the service phase becomes especially relevant and needs to be addressed in future research.
This paper has proposed a new, holistic and process-oriented categorization framework for ECM literature. This framework visualises the big picture of the ECM research field and allows a comprehensive coverage and precise categorization of publications in ECM and related areas. Drawing on an exhaustive systematic literature review which identified 384 journal articles and conference papers and 43 books, book sections, and other reports, the framework was used to categorize these publications and generate the current picture of research in ECM.
As depicted in Figure 7, this picture shows that major research areas are in (A) Prechange stage, product- and process-oriented research to reduce the impact of ECs, in (B) In-change stage, impact analysis, documentation and review, and systems to ease the handling with ECs, and in (C) Postchange stage, general surveys on sources and impacts of ECs to learn for future.
In contrast, little research has been done in (A) Prechange stage on people-oriented EC reduction measures, in (B) In-change stage on organizational issues and implementation of ECs, and in (C) Postchange stage on impact of ECs on quality and manufacturing & postmanufacturing stage. Subsequently, this paper discussed research opportunities to fill the observed gaps.
Moreover, the citation database Scopus was used to retrieve the citations of the publications categorized within the core ECM field, draw an h-graph, and analyze the links between the publications and between the categories using DSMs. The h-graph indicated 20 publications with 20 or more citations (h-index 20). The analysis of the category links showed that intracategory citations are usually higher but at the same time many categories are interlinked with intercategory citations. Overall, the citation analysis underlined that ECM is a relatively young and yet low-cited research field.
The result of this positioning and analysis can be used by both researchers and practitioners to (1) look for relevant publications for their research or work, (2) position their research or work in the overall picture of ECM, (3) focus on the identified research gaps and weak points, and (4) search for further research and improvement opportunities.
The authors express their gratitude to Andrew Flintham (Engineering Design Centre, University of Cambridge) for writing the program code for the analysis of citation links. The authors also thank the reviewers for their useful and constructive comments. This work is partly funded by a UK Engineering and Physical Sciences Research Council grant.
Bahram Hamraz received the Diploma degree in Industrial Engineering & Management from Hamburg University of Technology, Germany in 2007. He worked as a process management assistant for Daimler South East Asia in Singapore and as a management consultant for A.T. Kearney at their Berlin office in Germany. As a member of A.T. Kearney's operations practice, Bahram supported large international companies on their way to operational excellence in competencies like supply chain management, production, innovation, and complexity management. He is currently a Ph.D. Candidate at the Engineering Design Centre (EDC), Department of Engineering, University of Cambridge, UK. Bahram's research is supervised by Professor John Clarkson, Dr. Nicholas Caldwell, and Dr. David Wynn and is concerned with improving the understanding of implications and the management of engineering changes within complex products. Bahram is also supporting CRESCENDO (Collaborative and Robust Engineering using Simulation Capability Enabling Next Design Optimization)—an EU project involving aerospace companies across Europe. Furthermore, Bahram supervises undergraduate students in Operations Management at the Judge Business School in Cambridge.
Nicholas Caldwell received the B.A. degree in Computer Science from the University of Cambridge, UK in 1994, and the Ph.D. degree in Engineering (knowledge-based engineering for the scanning electron microscope) also from the University of Cambridge in 1998. He is a Fellow of the Royal Microscopical Society, a Member of the British Computer Society, a Chartered Engineer, and a Chartered IT Professional. Nicholas Caldwell has been a Research Associate at the Department of Engineering in the University of Cambridge since 1998. He has been actively involved in numerous research projects, including the EPSRC Knowledge and Information Management Through-Life Grand Challenge and the EPSRC/BAE Systems NECTISE projects, and is currently leading the Engineering Design Centre's contribution to the EU Framework 7 program CRESCENDO (Collaborative and Robust Engineering using Simulation Capability Enabling Next Design Optimization). He has also collaborated directly with BP and Carl Zeiss on specific projects on change management practice and innovative microscopy technologies. His research and technology interests include process modeling, engineering change management, knowledge management, electron microscopy, and multimedia design. He is actively involved in undergraduate engineering teaching and supervision of postgraduate students.
John Clarkson received the B.A. degree in Engineering (Electrical Sciences) from the University of Cambridge, UK in 1984 and the Ph.D. degree in Engineering (Electrical Machines) also from the University of Cambridge in 1988. In 2012, John received a Doctor Honoris Causa in Engineering Design from KU Leuven, Belgium. John Clarkson returned to the Department of Engineering of the University of Cambridge in 1995 following a 7-year spell with PA Consulting Group's Technology Division, where he was Manager of the Advanced Process Group. He was appointed director of the Engineering Design Centre (EDC) in 1997 and a University Professor in 2004. John is directly involved in the teaching of design at all levels of the undergraduate course. At PA John gained wide experience of product development with a particular focus on the design of medical equipment and high-integrity systems, where clients required a risk-based systems approach to design to ensure timely delivery of safe systems. His research interests are in the general area of engineering design, particularly the development of design methodologies to address specific design issues, for example, process management, change management, healthcare design, and inclusive design. As well as publishing over 450 papers, he has written and edited a number of books on medical equipment design and inclusive design.