Analyzing the influence factors of the post-earthquake reconstruction project using fuzzy DEMATEL

ABSTRACT

The high degree of integration of the Engineering-Procurement-Construction (EPC) method can more effectively reduce waste during construction and improve the sustainability of post-disaster reconstruction. The post-earthquake reconstruction in China is increasingly using the EPC method, and it is necessary to sort out the applicability and operating mechanism of the EPC method in the context of reconstruction. This study first established a process model for the post-earthquake reconstruction EPC project, and extracted 15 influencing factors. By using the fuzzy DEMATEL method, the prominence and mutual influence relationship between factors are determined. After setting a reasonable threshold, the key factors and influence paths are revealed. The study finds that the owner’s behavior, owner’s requirement, general contractor, available resources, contract and plan, and optimization information are the factors that need to be prioritized for resource allocating and managing, and they have a clear path of influence on other factors.

KEYWORDS:

• Post-earthquake reconstruction
• fuzzy DEMATEL
• project influence factor
• IDEF0

1. Introduction

Seismic activity wreaks havoc on infrastructure and inflicts economic harm on areas or nations; hence, it is vital to return infrastructure to normal through reconstruction. Reconstruction is defined as the actions to restore and improve pre-disaster living circumstances in disaster-affected communities (Hidayat and Egbu 2010)，and is a complex, multisectoral process requiring large resources and a wide range of expertise (Da Silva 2010). The reconstruction is not only the restoration of the original state, but also an opportunity to enhance the existing state for improved protection against future damage (Cheek and Chmutina 2022; Pribadi et al. 2014). Reconstruction includes society, economy, culture, and ecology in a broad sense, but these functions must be founded on the reconstruction of infrastructure and other physical systems. Therefore, a thorough understanding of the requirements and methods of infrastructure reconstruction is required. However, it must be noted that post-disaster reconstruction projects have unique characteristics that have changed both external constraints and internal working mechanisms, necessitating appropriate adjustments (Fayazi et al. 2017; Rouhanizadeh and Kermanshachi 2020; Vahanvati and Mulligan 2017). Due to suddenness, unclear planning, and multi-subject participation, the reconstruction exposed the lack of systematic arrangements for waste treatment and recycling (Pradhananga, Kasabdji, and Elzomor 2020), the insufficient constructability design (Yi and Yang 2014), and the existing construction methods cannot effectively support post-disaster reconstruction to maintain sustainability (Ismail, Halog, and Smith 2017).

The Engineering-procurement-construction (EPC) method has schedule and cost advantages over traditional methods (Franz, Molenaar, and Roberts 2020), hence China’s post-earthquake reconstruction projects are increasingly adopting the EPC method. However, the post-disaster reconstruction situation has exacerbated the complexity of EPC, and there is also a lack of success factors summarized from the EPC process.

The purpose of this study is to conduct a causal analysis of the influencing factors of post-earthquake reconstruction projects in China, so as to find an optimal construction path. Firstly, this article established the process model of China’s reconstruction EPC project, and identified a series of factors that affected the project from the model. Secondly, the fuzzy DEMATEL method was used to analyze the influencing and affected status of project influencing factors. Finally, by setting the threshold, the significant factors and interdependencies are screened, so as to determine the path of optimizing the reconstruction EPC project.

2. Background

2.1. Critical factor of post-disaster reconstruction

Post-disaster reconstruction projects can be considered as extensions of construction projects in the context of reconstruction, and the general principles and methods of projects are also applicable to reconstruction projects (Ismail et al. 2014; LaBrosse 2007). Compared with conventional projects, post-disaster reconstruction projects are more dynamic and complex, such as more unstable construction environment (Hidayat and Egbu 2010), a larger number of stakeholders that are more difficult to coordinate (LaBrosse 2007), and greater time pressure (Mannakkara, Wilkinson, and Potangaroa 2014).

The identification of project success factors has long been a focus of project management scholars’ research. Project success can be considered as achieving the needs of the project owner or users (Jugdev and Müller 2005). Critical success factors (CSF) for a project can be defined as factors that predict project success (Sanvido et al. 1992). If project participants are able to recognize CSF, they can avoid unsuccessful projects and increase the probability of project success. Research showed that it is impossible to create a general list of project success criteria suitable to all projects (Westerveld 2003), the judgments of project success may vary among stakeholders as it depends on the perspective of the evaluator (Lim and Mohamed 1999). Nevertheless, it is still necessary to develop a general set of criteria to cover the entire issue of project success and to assist project managers in their effective management activities.

Previous studies have investigated the critical success factors of post-disaster reconstruction projects from different angles. Some researchers positively considered project success factors or challenges based on project cases and literature (Ismail, Halog, and Smith 2017; Attalla, Hegazy, and Elbeltagi 2004; Da Silva 2010; Hidayat and Egbu 2011; Ophiyandri et al. 2013), others tried to find problems that projects should avoid from failure cases (Khalid et al. 2017; Sadiqi, Coffey, and Trigunarsyah 2011; Shaw, Gupta, and Sarma 2003), and some made retrospective summaries from the perspective of post-occupancy evaluation (Adamy and Bakar 2019).

However, these studies indicated that a successful reconstruction project should have these elements, which can be used as an indicator for judging the success of the project, but it is difficult to use this as a management tool in project management practice, and it is also challenging to arrange work and resources based on this. In addition, it is not true that most of the studies assumed that the key factors identified are not correlated or dependent on each other. Finally, in actual work, post-disaster reconstruction projects cannot simultaneously meet the dozens of performance evaluation indicators concluded in the research, and there must be some trade-offs. However, these researches did not provide a structured method to help decision-makers make reasonable trade-offs.

2.2. The process of China’s post-earthquake reconstruction project

Extracting success factors from completed projects will be disturbed by differences in environment, project type, etc. Therefore, this study employed an alternative method of finding influencing factors from the project process. Each project activity required input from the outside environment or other activities, and produced the output for other activities, meanwhile accepting support and constraints. The information on these four elements forms the necessary factors for the success of the project and has strong stability. Sanvido et al. (1992) developed the IBPM process model to identify key success factors for construction projects. Chan et al. (2005) analyzed BOT project factors in China. Papajohn et al. (2019) identified DB and CM contract management factors. Liu and Fang (2006) combined IDEF0 and Petri net for knowledge management and analysis of emergency tasks. Yung et al. (2014) illustrated the constructional coordination process of MEP in detail. Dachyar and Sanjiwo (2018) analyzed the factors and reengineer the business process reengineering in the oil and gas EPC projects.

In the EPC process, the owner typically contracts with a single general contractor (Du et al. 2016; Shen et al. 2017). The contractor assumes full responsibility for the engineering, procurement, construction, commissioning, and final turnover of a complete facility to the owner (FIDIC 2017; Pal, Wang, and Liang 2017). Under a single contract, the contractor has greater initiative and motivation to adjust the work of each stage (Pham and Hadikusumo 2014), reducing the waste of resources caused by poor cooperation between different departments. Shortening project duration usually relies on overlapping activities, which increases the interdependence and uncertainty of activities, increases the difficulty of contractor work decomposition and planning, and requires a more complex organizational structure to adapt (Yeo and Ning 2002).

Although the FIDIC drafters suggested that the EPC method should not be used when there is insufficient time or information for risk assessment or when there are a large number of areas that bidders cannot inspect (FIDIC 2017), the time advantage of EPC still attracted owners of post-earthquake reconstruction projects in China. China’s regulations, the construction market, and the characteristics of reconstruction projects have modified the EPC procedure. Therefore, this study is based on Chinese laws and regulations, the EPC process model summarized in previous studies (Dachyar and Sanjiwo 2018; Karhu 2001; Papajohn 2019; Sanvido and Khayyal 1990; Yeo and Ning 2002), and a survey of project participants, with the help of the IDEF0 tool to establish a reconstruction EPC process model suitable for China, as shown in Figure 1. The boxes represent the main activities(functions)in the EPC process. The connecting arrows between the main functions in the model represent the preconditions of each stage or work in the construction process, the output to other work, the constraints. These four groups of information actually constitute the factors that affect the smooth development of each work. After merging some factors according to the similarity of management methods, 15 project success factors are identified, as shown in Table 1.

Figure 1. China’s post-earthquake reconstruction EPC process model.

Table 1. The project success factors extracted from the process model.

Factor	Code	Explanation
Site and facility	SF	Including the site’s geological conditions, topography, site scale, shape, connection status between the site and the outside, the status of existing facilities in the site, and other materials and their representative information.
Available resource	AR	It includes the resources that all parties put into the project to transform or consume to form project products.
Owner’s behavior	OH	The owner’s behavior refers to various behaviors such as supervision, support, and intervention of the owner (including the consulting company entrusted by him) to the project and the contractor during the entire construction period.
Owner’s requirement	OQ	The owner’s requirements are various quantifiable requirements for the project generated by the owner, including the function, scale, and coordination of the project.
Contract and plan	CP	Contracts and plans regulate the behavior of all parties and stipulate the methods and procedures for handling various matters.
Project experience	PE	Project experience is the information and knowledge gradually generated and accumulated in the whole process of construction, and it reacts with various activities of the project.
General contractor	GC	The general contractor includes the project execution team (EPC management team, design team, procurement team, construction team) and collaboration departments (bidding department, financial department, logistics support department).
Method and tool	MT	Methods and tools refer to various technical methods and tools used to improve performance, including professional technology and management technology.
Participant constraint	CC	Participant constraints are mainly the rules and management systems of participating companies and those established for the project, which have always imposed constraints on the project from the preparation stage to the completion of the project construction.
External constraint	EC	External factors refer to the external environmental factors of the project that may hinder or affect the construction, including the natural environment, laws, regulations, norms and government permits, social culture, ethnic and religious habits, industry and market trading habits, logistics and transportation conditions, etc.
Project product	PP	The physical outcomes of the project at various stages in the process.
project Idea	PO	The project concept is the starting point of project construction, which stipulates the basic concepts or ideas such as the type of project, main objectives, main functions to be realized, and scale, and gradually forms implementable project requirements after the preparation stage
Technical document	TD	Technical documents refer to the data that provide technical guidance to project personnel and monitor the project during the construction process.
Optimization information	OI	Optimization information includes constructability information and changes, construction technical information, and material and equipment technical information. This information is feedback from subsequent steps to optimize the output of previous stages.
Performance information	PI	Performance indicators include progress, cost, quality, customer satisfaction, etc.

3. Research method

3.1. Methodology

The success factors of the project can be identified through the process model; however, it is still difficult to judge the relationship between the factors, and there is no quantitative support for understanding the criticality of the factors. Therefore, this section discusses how to quantitatively judge the importance of factors and the interdependence among them.

DEMATEL（Decision Making Trial and Evaluation Laboratory）is a systematic method of factor analysis (Si et al. 2018). It is often used in construction engineering to analyze various elements and components interacting in project systems (Han and Wang 2018; Mavi and Standing 2018; Shakeri and Khalilzadeh 2020) and is suitable for the analysis of key factors and their mutual influence in post-disaster reconstruction EPC projects. The technique comprehensively considers the status of factors in the system and the interaction between factors. DEMATEL uses graph theory and matrix tools to construct a judgment matrix to analyze system elements and visualize complex causal relationship structures. After quantifying the dependence of indicators, by calculating the degree of mutual influence of each factor, as well as the prominence and cause value, the causal relationship between factors and the status of factors in the system is displayed (Fontela and Gabus 1974). DEMATEL and derived fuzzy DEMATEL, grey DEMATEL are used in many fields to determine critical factors (Chang, Chang, and Wu 2011; Mahmoudi et al. 2019; Mavi and Standing 2018; Raval, Kant, and Shankar 2021). After calculation, four types of data can be obtained, that is “Influence degree”(D), “affected degree”(R), “Prominence”(D+R), and “Relation”(D-R). D indicates direct and indirect effects given by a factor to the other factors， R indicates direct and indirect effects by a factor from the other factors, (D+R) implies the strength of influences that are given and received by the factor, and (D-R) shows the net effect that factor contributes to the system (Si et al. 2018; Tseng 2009). When (D-R) is positive, the factors belong to the cause group and have a net influence on the other factors; and (D-R) is negative, the factors are the effect group and influenced by the other factors. The (D+R) and (D-R) of all factors are drawn in a causal diagram to assist in judging the importance (Mahmoudi et al. 2019; Patil and Kant 2013; Zhou, Huang, and Zhang 2011; Zhou et al. 2017, 2018). Finally, by setting a reasonable threshold, insignificant factors and dependencies can be eliminated (Feng and Ma 2020), and managers can identify paths for optimization or improvement in a more efficient manner.

By calculating the mean of the (D+R), the factors can be divided into four quadrants (Chien, Wu, and Huang 2014; Chuang et al. 2013; Hwang et al. 2016). Factors’ prominence greater than the mean value and with positive (D-R) are core factors with high significance and relationship; prominence smaller than the mean and with positive (D-R) are driving factors; factors less than the mean and negative (D-R) have low prominence and relation, is separated from the system and are the independent factors; factors greater than the mean and with negative (D-R) are impact factors which affected by other factors and cannot be directly improved (Si et al. 2018).

Because practical problems are often complex, the degree of influence among various factors is uncertain, and individual experts have different understandings of problems. Therefore, Experts cannot use precise values to judge the degree of influence (Herrera-Viedma et al. 2005; Wang and Chen 2005). Fuzzy DEMATEL introduces fuzzy sets to deal with the ambiguity of natural language expression and adjust respondents’ judgments on the relationship between factors. The method improves the accuracy of the DEMATEL by pre-setting the corresponding relationship between natural language such as “weak influence”, “strong influence”, etc. and triangular fuzzy numbers, and then obtains clear values through defuzzification (Si et al. 2018)。

The steps of the fuzzy DEMATEL are as follows

Step 1: Generating the direct-relation matrix. The pair-wise comparison method was used in the questionnaire to obtain respondents’ judgments on the effects and direction between factors, and a five-level semantic evaluation structure was designed, such as “no influence”, “very weak influence”, “weak influence”, “strong influence”, “very strong influence”, and correspond the semantic expression to the triangular fuzzy number were shown in Table 2.

Table 2. Semantic evaluation and the triangular fuzzy numbers.

Linguistics assessment	Influence score	Triangular fuzzy numbers
No influence	0	（0.00，0.00，0.25）
Very weak influence	1	（0.00，0.25，0.50）
weak influence	2	（0.25，0.50，0.75）
strong influence	3	（0.50，0.75，1.00）
Very strong influence	4	（0.75，1.00，1.00）

Step 2: Normalizing the direct-relation matrix. Following steps (1)–(5) below, converting fuzzy data into crisp scores used the defuzzification method (Opricovic and Tzeng 2003).

(1) Normalization:

(1) $\mathit{x}{r}_{ij}^n=\left(r_{ij}^n-min{l}_{ij}^n\right)/{\mathrm{\Delta }}_{min}^{max}$(1)

(2) $\mathit{x}{m}_{ij}^n=\left(m_{ij}^n-min{l}_{ij}^n\right)/{\mathrm{\Delta }}_{min}^{max}$(2)

(3) $\mathit{x}{l}_{ij}^n=\left(l_{ij}^n-min{l}_{ij}^n\right)/{\mathrm{\Delta }}_{min}^{max}$(3)

Where ${\mathrm{\Delta }}_{min}^{max}=max{r}_{ij}^n-min{l}_{ij}^n$

(2): Compute the left (ls) and right (rs) normalized values:

(4) $\mathit{xr}{s}_{ij}^n=\mathit{x}{r}_{ij}^n/\left(1+\mathit{x}{r}_{ij}^n-\mathit{x}{m}_{ij}^n\right)$(4)

(5) $\mathit{xl}{s}_{ij}^n=\mathit{x}{m}_{ij}^n/\left(1+\mathit{x}{m}_{ij}^n-\mathit{x}{l}_{ij}^n\right)$(5)

(3): Compute total normalized crisp values:

(6) $x_{ij}^n=\left[\mathit{xl}{s}_{ij}^n\left(1-\mathit{xl}{s}_{ij}^n\right)+\mathit{xr}{s}_{ij}^n\times \mathit{xr}{s}_{ij}^n\right]/\left[1-\mathit{xl}{s}_{ij}^n+\mathit{xr}{s}_{ij}^n\right]$(6)

(4): Compute crisp values:

(7) $z_{ij}^n=min{l}_{ij}^n+x_{ij}^n\times {\mathrm{\Delta }}_{min}^{max}$(7)

(5): Integrate crisp values:

(8) $z_{\mathit{ij}}=\frac{1}{p}\left(z_{ij}^1+z_{ij}^2+\cdots +z_{ij}^p\right)$(8)

where，p is the number of respondents.

Step 3: Attaining the total-relation matrix T using equations (9–11):

(9) $\mathit{M}=\underset{1\le \mathit{i}\le \mathit{n}}{max}{\sum }_{j=1}^n{z}_{\mathit{ij}}$(9)

(10) $\mathit{N}=\frac{Z}{M}$(10)

(11) $\mathit{T}=\mathit{N}{\left(1-\mathit{N}\right)}^{-1}$(11)

Step 4: Producing a causal diagram. The sum of rows and the sum of the columns are D_i and R_i through equations (12–13). Then take (D+R) as the horizontal axis and (D-R) as the vertical axis, and draw the causal diagram.

(12) $D_i={\sum }_{j=1}^n{t}_{\mathit{ij}}$(12)

(13) $R_i={\sum }_{i-1}^n{t}_{\mathit{ij}}$(13)

3.2. Data collection and calculation

Based on the influencing factors extracted from the EPC process model (see Figure 1 and Table 1), and the steps of the fuzzy DEMATEL method, this study collected first-hand data from practice experts and professional scholars involved in China’s reconstruction projects through questionnaires. The questionnaire explained the meaning of each factor, and invited experts to judge whether there is an influence relationship among the factors and the degree of influence. Questionnaires were collected by email, interviews and fieldwork. Considering that the EPC method spans multiple departments and disciplines of owners, contractors, and suppliers, the respondents of the questionnaire are middle and senior managers of enterprises and projects who are familiar with the EPC management process, as well as academic experts who are interested in post-disaster reconstruction. These experts have participated in many post-earthquake reconstruction projects in China (including the Ya’an earthquake reconstruction in 2013, the Jiuzhaigou earthquake reconstruction in 2017, and the Yibin earthquake reconstruction in 2019) to ensure the high reliability of the sample data. A total of 30 questionnaires were distributed in this study, and 16 valid questionnaires were finally retained.

The original data obtained by experts’ investigation was transformed into triangular fuzzy numbers, and the initial direct relationship matrix was calculated. This article only listed the original data of one expert (see Table 3), converted it according to the corresponding relationship between semantic evaluation and triangular fuzzy numbers in Table 2, and calculated the initial direct relationship matrix of all factors (see Table 4) according to the Equations (1(1) $\mathit{x}{r}_{ij}^n=\left(r_{ij}^n-min{l}_{ij}^n\right)/{\mathrm{\Delta }}_{min}^{max}$(1) -8(8) $z_{\mathit{ij}}=\frac{1}{p}\left(z_{ij}^1+z_{ij}^2+\cdots +z_{ij}^p\right)$(8) ), Using the data in Table 4 with Equations (9(9) $\mathit{M}=\underset{1\le \mathit{i}\le \mathit{n}}{max}{\sum }_{j=1}^n{z}_{\mathit{ij}}$(9) -11(11) $\mathit{T}=\mathit{N}{\left(1-\mathit{N}\right)}^{-1}$(11) ), the total relation matrix of the factors was finally obtained. The results are shown in Table 5.

Table 3. An example of original data from one expert.

	SF	AR	OH	OQ	CP	PE	GC	MT	CC	EC	PP	PO	TD	OI	PI
SF	0	0	3	0	0	0	3	3	1	4	2	3	2	4	1
AR	3	0	4	1	2	1	4	4	4	4	3	4	4	4	3
OH	1	0	0	0	0	0	2	1	0	3	0	2	2	1	1
OQ	2	2	4	0	2	1	4	4	3	4	2	4	4	4	3
CP	4	1	4	3	0	2	4	4	2	4	3	4	4	4	3
PE	3	2	4	1	2	0	4	4	3	4	2	4	4	4	3
GC	1	0	2	0	0	0	0	1	0	2	0	2	1	2	1
MT	1	0	2	0	0	0	1	0	1	3	0	2	1	2	1
CC	1	0	4	1	0	0	4	4	0	4	0	4	4	3	2
EC	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
PP	2	1	4	0	0	0	4	3	2	4	0	4	4	3	3
PO	0	0	2	0	0	0	2	2	1	3	0	0	1	1	1
TD	0	0	1	0	0	0	2	1	0	3	0	1	0	1	0
OI	1	0	1	0	0	0	1	1	1	4	1	2	1	0	0
PI	2	0	3	1	0	0	3	2	0	4	0	3	3	3	0

Table 4. The initial direct-relation matrix.

	SF	AR	OH	OQ	CP	PE	GC	MT	CC	EC	PP	PO	TD	OI	PI
SF	0.000	0.050	0.031	0.036	0.047	0.050	0.052	0.045	0.052	0.074	0.050	0.059	0.066	0.049	0.052
AR	0.054	0.000	0.038	0.052	0.057	0.063	0.054	0.054	0.061	0.077	0.068	0.072	0.070	0.066	0.057
OH	0.072	0.057	0.000	0.061	0.065	0.063	0.063	0.070	0.074	0.091	0.079	0.072	0.079	0.072	0.083
OQ	0.070	0.061	0.043	0.000	0.064	0.068	0.066	0.061	0.061	0.086	0.070	0.077	0.081	0.065	0.072
CP	0.068	0.047	0.043	0.052	0.000	0.057	0.050	0.049	0.056	0.079	0.065	0.066	0.081	0.054	0.061
PE	0.052	0.040	0.036	0.042	0.040	0.000	0.042	0.047	0.056	0.074	0.059	0.072	0.077	0.049	0.045
GC	0.058	0.054	0.040	0.054	0.052	0.054	0.000	0.056	0.065	0.077	0.059	0.066	0.072	0.061	0.065
MT	0.058	0.043	0.035	0.051	0.047	0.061	0.040	0.000	0.061	0.074	0.063	0.061	0.074	0.045	0.058
CC	0.047	0.043	0.036	0.033	0.040	0.036	0.049	0.038	0.000	0.063	0.050	0.049	0.066	0.047	0.052
EC	0.024	0.026	0.010	0.017	0.024	0.027	0.026	0.020	0.031	0.000	0.029	0.029	0.033	0.026	0.035
PP	0.052	0.040	0.028	0.035	0.047	0.038	0.042	0.043	0.052	0.066	0.000	0.056	0.061	0.052	0.049
PO	0.043	0.040	0.038	0.026	0.036	0.041	0.043	0.040	0.045	0.066	0.047	0.000	0.056	0.038	0.043
TD	0.034	0.033	0.017	0.031	0.034	0.034	0.027	0.026	0.038	0.052	0.036	0.047	0.000	0.040	0.036
OI	0.047	0.040	0.031	0.042	0.040	0.054	0.052	0.059	0.061	0.079	0.061	0.061	0.082	0.000	0.061
PI	0.045	0.036	0.026	0.038	0.033	0.032	0.036	0.049	0.050	0.072	0.050	0.058	0.056	0.049	0.000

Table 5. The total relation matrix of each factor.

	SF	AR	OH	OQ	CP	PE	GC	MT	CC	EC	PP	PO	TD	OI	PI
SF	0.109	0.141	0.099	0.121	0.140	0.151	0.147	0.142	0.165	0.224	0.166	0.183	0.205	0.154	0.165
AR	0.180	0.110	0.119	0.151	0.166	0.181	0.166	0.168	0.194	0.255	0.203	0.218	0.235	0.189	0.191
OH	0.219	0.185	0.097	0.178	0.193	0.202	0.195	0.204	0.230	0.301	0.238	0.244	0.273	0.217	0.238
OQ	0.209	0.181	0.133	0.113	0.186	0.199	0.191	0.188	0.210	0.285	0.221	0.240	0.264	0.202	0.220
CP	0.190	0.152	0.121	0.149	0.110	0.173	0.160	0.160	0.186	0.253	0.197	0.209	0.240	0.175	0.191
PE	0.159	0.133	0.105	0.127	0.134	0.104	0.138	0.145	0.169	0.226	0.175	0.196	0.217	0.155	0.160
GC	0.181	0.160	0.119	0.151	0.160	0.171	0.114	0.168	0.195	0.253	0.193	0.210	0.234	0.183	0.196
MT	0.171	0.141	0.108	0.140	0.146	0.168	0.143	0.106	0.181	0.235	0.186	0.194	0.222	0.158	0.178
CC	0.144	0.127	0.098	0.110	0.125	0.129	0.135	0.127	0.105	0.201	0.156	0.163	0.193	0.143	0.156
EC	0.079	0.072	0.045	0.060	0.071	0.079	0.074	0.070	0.089	0.079	0.088	0.093	0.104	0.080	0.092
PP	0.150	0.125	0.091	0.113	0.132	0.131	0.130	0.133	0.156	0.206	0.109	0.170	0.190	0.149	0.153
PO	0.134	0.118	0.095	0.098	0.116	0.127	0.124	0.122	0.141	0.194	0.145	0.108	0.174	0.128	0.140
TD	0.108	0.096	0.065	0.089	0.099	0.104	0.094	0.093	0.116	0.157	0.116	0.132	0.099	0.112	0.115
OI	0.160	0.137	0.104	0.131	0.139	0.160	0.153	0.161	0.180	0.239	0.183	0.193	0.229	0.114	0.180
PI	0.138	0.117	0.086	0.112	0.115	0.122	0.120	0.133	0.149	0.204	0.151	0.166	0.178	0.140	0.101

Note: Text in bold indicates a significant relationship path.

According to the Equations (12)(12) $D_i={\sum }_{j=1}^n{t}_{\mathit{ij}}$(12) and (13(13) $R_i={\sum }_{i-1}^n{t}_{\mathit{ij}}$(13) ), the influence degree D, the affected degree R, prominence (D+R), relation (D-R) of each factor is obtained and listed in Table 6. The ordering of each factor under the different was after the numerical value. According to the (D-R), factors were divided into net cause and net effect categories.

Table 6. The cause and effect values.

code	Influence factor	D	R	D+R	D-R	Category
SF	Site and facility	2.311(9)	2.331(7)	4.642(10)	−0.020(9)	Net effect
AR	Available resource	2.726(3)	1.996(13)	4.722(4)	0.731(3)	Net cause
OH	Owner’s behavior	3.214(1)	1.485(15)	4.699(5)	1.729(1)	Net cause
OQ	Owner’s requirement	3.041(2)	1.843(14)	4.884(1)	1.198(2)	Net cause
CP	Contract and plan	2.665(5)	2.032(12)	4.697(6)	0.633(4)	Net cause
PE	Project experience	2.342(8)	2.199(9)	4.541(13)	0.143(8)	Net cause
GC	General contractor	2.688(4)	2.085(11)	4.773(2)	0.603(5)	Net cause
MT	Method and tool	2.476(6)	2.120(10)	4.595(11)	0.356(6)	Net cause
CC	Participant constraint	2.114(11)	2.465(6)	4.579(12)	−0.351(10)	Net effect
EC	External constraint	1.177(15)	3.312(1)	4.488(15)	−2.135(15)	Net effect
PP	Project product	2.138(10)	2.529(4)	4.667(8)	−0.390(11)	Net effect
PO	Project Idea	1.964(13)	2.718(3)	4.682(6)	−0.754(13)	Net effect
TD	Technical document	1.595(14)	3.059(2)	4.654(9)	−1.463(14)	Net effect
OI	Optimization information	2.461(7)	2.298(8)	4.760(3)	0.163(7)	Net effect
PI	Performance information	2.033(12)	2.475(5)	4.508(14)	−0.443(12)	Net effect

4. Results and analysis

The results in Figure 2 showed that the prominence values of the factors of the reconstruction EPC project are all positive, indicating that each factor plays an important role in the reconstruction of the EPC project. At the same time, the prominence distribution interval of all factors is [4.488, 4.884], indicating that the importance of each factor in management has little difference. Therefore, the determination of the key factors is mainly carried out according to the degree of cause. The mean of the prominence is 4.659, and the factors are divided into four quadrants based on the mean value.

Figure 2. the causal-effect diagram.

Eight factors were positive and divided into the cause group, including OH, OQ, AR, CP, GC, MT, OI and PE, while the effect group was composed of such factors as EC, TD, PO, PI, PP, CC, SF. Generally, the cause group factors are difficult to move, while the effect group factors are easily moved (Tseng 2009). Therefore, when making decisions or optimizing projects, managers should focus on formulating measures from the cause group factors.

Further, six factors were grouped into the core factors including OH, OQ, AR, CP, GC and OI. They have high prominence and high relation and should be received the highest priority in terms of management resources. PE and MT are causal factors as well as driving factors. SF, CC, PI, TD and EC are classified into the independent factor group. Finally, PP and PO are the impact factors. By acquiring this information, the resource allocation sequence should be from core factor to driving factor, then to the independent factor, and finally to the impact factor (Chien, Wu, and Huang 2014).

Through Tables 6 and Figure 2, we can find the cause factors and effect factors influencing the post-earthquake reconstruction EPC project in China. However, the interrelationship of these factors cannot be reflected by causal-effect diagrams, nor can these relationships be exploited in future optimizations. To solve this problem, the article sets an appropriate threshold for the comprehensive impact matrix to obtain more structured analysis results. Obviously, if the threshold is too low, the retained factors and their relationships will be too complex, reducing the efficiency and effectiveness of decision-making; but if the threshold is too high, effective factors and relationships will be deleted. The threshold usually is determined by expert discussion and brainstorming (Si et al. 2018), the maximum mean de-entropy (Li and Tzeng 2009), the average of all elements in the matrix T (Sara, Stikkelman, and Herder 2015), the maximum value of the diagonal elements of the matrix T (Wee-Kheng and Kuo 2014) or mean $\pm$ standard deviation (Feng and Ma 2020), etc. The threshold value in this article is determined by the change in the number of relationships between factors. Using 1/10 of the difference between the maximum value and the minimum value among all factors as the common difference, increasing from the minimum value in turn, and calculating the number of factor relationships exceeding the threshold. Figure 3 demonstrated that when the cumulative value increases, there are two distinct inflection points in the change in the number of relationships. The value corresponding to the second inflection point was taken as the threshold value, at which point the number of interdependencies between factors changed from a rapid decrease to a relatively slow decrease. After a comprehensive analysis, the threshold value was set at 0.2244, and 19 relationships between factors were remained. Combined with the relationship shown in Table 5, we can draw a relationship path diagram shown in Figure 4. The coefficients of each path in Figure 4 indicate the importance of the interaction between factors, and also indicate the difference in the effect of factor changes during optimization.

Figure 3. The amount of relationship with different threshold.

Figure 4. The impact-relation map based on the threshold value 0.2244.

5. Discussion

In recent years, the occurrence of disasters and damage have increased. The particularity of post-disaster reconstruction requires decision-makers and managers to optimize the construction process. From the perspective of active control, although some of the affected factors are important to the project, the decision maker should choose the factors that have a net impact on other factors to optimize. Six factors including owner’s behavior and owner’s demand, resource availability, contractor’s ability, contract and plan, and timely and efficient optimization information should be taken as the main optimization elements.

The results in Figures 2 and 4 showed that the owner (including behavior and requirement) is the most important among all factors, and nine of the 19 significant influence paths are related to it indicated the owner provides the broadest influence on the reconstruction project. The two factors directly affected performance information (PI), project products (PP), project concept (PO), participant constraint (CC), external constraints (EC) and project technical documents (TD), which run through the whole process of the project. Among them, the owner’s stable and clear requirements could reduce the uncertainty of project delivery and reduce rework and overruns (Okada, Simons, and Sattineni 2017); the owner’s behaviors of authorization scope for the project, the degree and method of intervention in the project directly changed the project process. Therefore, the owner needs to innovate its management method to obtain the maximum performance. Although most disasters are sudden and indiscriminate and local managers cannot make detailed and clear reconstruction plans for specific facilities, but it is possible to determine the requirements of reconstruction projects based on local overall development plans and goals and maintain stability.

The (D-R) values of general contractor (GC), available resource (AR) and contract and plan (CP) have little difference, and the influences on EC and TD respectively are roughly close. These three factors themselves have a strong correlation and can actually be considered as the general contractor’s performance in terms of management behavior, resource integration and planning. In view of the particularity of post-disaster reconstruction in China, the general contractors of reconstruction projects are all state-owned enterprises, and their responsibilities far exceed those of ordinary commercial projects contractors. This requires that the general contractor must exert influence on other factors and put them under its control in order to ensure the completion of the project objectives. Resources such as materials and equipment constitute the important basis of the project, and their quality, quantity, and entry time are all rigid constraints. The availability of resources (including timeliness, stability, type, quantity and quality) has a greater impact on the project. The deployment of subcontractors and work teams, the use of methods and tools, and plan optimization all need to be adjusted in real time according to resource conditions. Obviously, contract and plan (CP) is a code of behavior signed by the owner and the contractor, the supplier. The direction of optimization should be based on the inspection of the contractor’s technical capabilities, incident response capabilities, resource guarantee capabilities and the stability of the contractor.

The (D-R) values of method and tool (MT), project experience (PE), and optimization information (OI) were positive. Appropriate methods and tools help to promote project faster and better. Usually, the design methods and tools are restricted by the ability of the designer, and the construction methods and tools are determined according to the design requirements. However, in the reconstruction environment, it is not that advanced technologies are more popular, On the contrary, in order to reduce the risks associated with advanced methods, all participants emphasize familiarity, stability and popularity of methods and tools. PE and OI were constantly enriched and adjusted during the project progress. From the perspective of the process, the project experience and optimization information formed at any stage will have a positive effect on the follow-up work and play an indispensable role in reducing project deviation. construction behavior is more pronounced. OI was grouped into core element, indicating that it is necessary to focus on allocating resources to improve the efficiency and effectiveness of project participants in delivering optimization information. In short, the closer these three types of information are to the reality of the reconstruction project, the more efficient the project’s transformation, conversion, value-added and communication can be improved.

In Table 2, the external constraint (EC) ranked 15th in (D-R), indicating that this factor was most easily affected by other factors. Previous studies have argued that external factors will affect the success of a construction project (Chan, Scott, and Chan 2004). However, in the reconstruction scenario in China, the proactive efforts of owners and contractors will have a positive impact and convert it into factors that promote the project, as shown in the eight impact paths in Figure 4. When active management measures are adopted, except for the hard constraints such as quality and safety, many external constraints can be transformed into positive driving factors, such as local culture and religious customs in external constraints. Designers have an in-depth understanding of local characteristics in advance, in-depth study of ethnic and religious policies, listen to the opinions of local people during the design process, and fully integrate local cultural and religious characteristics into the design; appropriately absorb some local people to participate in the construction. These practices can minimize the negative impact of culture and religion on the project, and may promote the smooth implementation of the project, and these adjustments are precisely caused by the improvement of the corresponding cause factors. It is worth noting that the owner’s influence coefficient on external constraint is the largest among all coefficients (0.3011 and 0.2845 respectively in Figure 4), which showed that the owner plays an important role in coordinating the external stakeholders of the project. Therefore, when optimizing, measures should be taken with the goal of identifying and stabilizing demand early and giving contractors more assistance rather than intervention.

The technical document (TD) has a positive effect on the project (D+R value is 4.654). It is not only the summary of the project results but also the guidance of the project implementation. The project manager accurately evaluates the project status through the technical document and adjusts the management method in time and the project personnel follows the technical documents to complete the project. After filtering with a threshold of 0.2244, there are 6 significant influence paths from the owner and the general contractor exerting influence on it. The technical documents summarize the information of various results produced by the joint work of the owner and the contractor, and truly reflect the formation process and performance of the project, as well as the control of the project. The environmental changes of the reconstruction project are frequent and drastic, and the technical documents summed up in time by the owner and the contractor can minimize the blindness and confusion of subcontractors and workers. Due to the characteristics of rapidity and environmental uncertainty of reconstruction projects (Hidayat and Egbu 2010; Mannakkara, Wilkinson, and Potangaroa 2014), performance information (PI), project product (PP), and project idea (PO) generally cannot directly refer to the goals and requirements of commercial projects need to rely more on the definition or setting of the owner. Therefore, the behavior and needs of the owner have great influence on this Three factors produced a more significant impact path. Participants are entrusted by the owner, so their constraints on the project (CC) are directly affected by the owner’s behavior.

6. Conclusion

China’s post-disaster reconstruction projects are gradually adopting the EPC method, but the characteristics of the project determine that it is necessary to adapt and optimize the EPC process. Decision-making and optimization need to take into account the factors that affect each other and the dependencies between them, find out the factors that have more influence on other factors, and improve the system along the main impact path. This study established a process model suitable for the characteristics of China’s post-earthquake reconstruction EPC projects and identified 15 influencing factors from it. The influencing factors identified from the project process exclude the interference of geography, project type and others to the greatest extent.

Using the fuzzy DEMATEL method, the 15 factors are divided into the cause group factors and effect group factors. Although each factor will interact with the other, by setting an appropriate threshold, some relations and paths with weak dependencies can be deleted, and the influence between important factors is more clearly displayed. The results show that despite the adoption of the EPC method, the owner is still the most important party in China’s post-earthquake reconstruction projects. Owners need to take steps to identify the clear requirements early and maintain it stable, while providing more support and less intervention to contractors. As the main participant and responsible party of EPC, the general contractor should have sufficient ability to integrate factors such as available resources and contract and plans, and strengthen the efficiency and effect of optimization information. After a disaster, resources should be allocated more towards these six factors to improve reconstruction performance. When optimizing the management of projects, the effect of optimization measures can be evaluated more clearly based on the path coefficient, so that decisions can be made more accurately.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

QingPing Zhong

QingPing Zhong, Ph.D. Research field: project cost management, engineering management in disaster.

Hui Tang

Hui Tang, Ph.D. Research field: project cost management, engineering disaster mitigation.

Wenmei Zhou

Wenmei Zhou, Ph.D. Research field: disaster emergency management.