Cause analysis of construction safety accidents in China using association rules

Abstract

In China the number of accidents and deaths in construction industry is increasing rapidly in recent years. Cause analysis is important to reveal the main factors contributing to the accidents, and therefore improve the safety mechanism for accident reduction in future. In this paper, 653 construction safety accidents in China which are fully investigated and formulated by a set of country-specific factors are used as the source data of potential causal relationships among the factors. By utilizing association rules a total of 13 main factors contributing to accident causation are explored, and 9 factors significantly impacting on the severity of accidents are identified. The discovered association rules offer scientific grounds for Chinese construction safety management and decision support. The novelty of this study is to present a different perspective during cause analysis in which the factors impacting the construction site safety and the consequences of construction accidents are revealed respectively.

Keywords

Cause analysis construction safety accident association rule lift

1. Introduction

Figure 1.

The statistics of construction safety accidents in China from 2004 to 2019.

Construction is one of the high-risk industries in the world with frequently occurred safety accidents. Construction safety accidents are the abrupt events in building engineering activities which injure personal safety and health, damage equipment and facilities, cause economic losses, and result in a temporary suspension or even termination of construction activities. In the past years, around 340 million occupational accidents occurred annually resulting in 2.3 million injuries and 6000 deaths daily all around the world [3]. In China construction industry has become the pillar industry of national economic development at the cost of numerous casualties. The number of accidents and deaths in construction industry increases rapidly ranking the first in all fields of industrial production since 2012. Figure 1 shows the number of construction accidents and deaths in China from 2004 to 2019. Since the release of regulations on safety production management of construction projects in 2004, construction enterprises began to pay attention to safety management so that the number of construction accidents and deaths gradually declined with a small range of fluctuations. However the absolute number of accidents and deaths per year remains at a high level and rises rapidly since 2015, reaching 773 construction accidents and 904 deaths in 2019. Some typical constriction safety accidents are listed in Table 1. The high incidence of construction safety accidents leads to a large quantity of casualties and property losses due to the concentrated hazard sources, large proportion of migrant workers in low education level, frequent change of on-site workers, etc. It is therefore a vital issue of China to enhance the safety situation of construction industry and improve the safety mechanism of construction so as to promote the stable improvement of construction safety production.

According to accident cause theory, accident is the product of factors under the action of catalysis and criticality. Cause analysis which focuses on extracting the accident mechanism from the analysis of typical accidents, identifying the critical causing factors contributing to accidents, reducing the accident rate, eliminating the potential danger, ensuring the health and safety of employees remains a topic of significant importance in construction industry.

In the literature Heinrich noted that human unsafe behaviors, unsafe working conditions and equipment were important causes of industrial accidents [10]. Haslam regarded workers fault, workplace problems, PPE defects, construction materials, and defects in risk management as common factors of construction accidents. Other causing factors include emergency management factors (e.g., insufficient protection equipment, incomplete accident prevention plan) [3, 16], worker factors (e.g., unqualified worker, noncompliance behavior, inadequate execution of construction plan) [22], safety inspection factors (e.g., insufficient safety supervision, safety inspection, risk identification ability, and safety training) [7, 20], project planning factors (e.g., unqualified contractor, multi-level subcontractors) [9, 19], etc.

A diversity of methods were proposed to analyze the causing factors of accidents, such as Domino Model, HFACS (Human Factors Analysis and Classification System), STAMP (Systems Theoretic Accident Model and Processes), Haddon, TMAR (Theoretical Model of Accident Reaction), FTA (Fault Tree Analysis), Delphi, safety risk network [27, 29, 33]. In [15] the contribution of construction project features to accident causation was investigated including project nature, construction method, site restriction, project duration, procurement system, design complexity, level of construction, and subcontracting. Zhang et al. evaluated the applicability of four contemporary accident causation models, namely, STAMP, AcciMap, HFACS, and 2–4 Model [30] on a typical construction accident. Alomari et al. investigated the key factors including site conditions, human, environmental, social, economic, organizational, regulatory, psychological, and ergonomic factors contributing to construction worker safety using Delphi technique [2]. In [31], 39 factors on construction accidents are evaluated by grey relational analysis method and classified into three levels: critical, important, and ordinary. However the relationship among these factors hidden the accidents was not investigated in the study.

In fact the construction accidents are usually caused by the violation of safety restrictions [13] which may trigger other factors [18]. Association analysis is intended to explore the possible relationship among objects from a large quantity of data. In a broad sense, the relationship can be concurrent relationship, causal relationship, temporal relationship etc. Association rules initially proposed to study the goods that are often purchased at the same time in the customer transaction database [1] has be broadly applied in diverse fields [6, 12]. Association rules able to discover the hidden relationship or patterns from data are found effective and prospective in cause analysis of various accidents, including railway accidents [17], construction accidents [23], marine accidents [32], fire accidents [11], work zone crash casualties [26], underground coal mine accidents [21], and elevator accidents [24], etc.

Table 1

Some construction safety accidents in China

Day	Location	Type	Loss
2010.11.15	Jiaozhou Road, Shanghai	Fire accident caused by illegal behavior	58 deaths and more than 70 injures
2011.6.19	Wuxi, Jiangsu Province	Collapse of construction site	9 deaths and 7 injuries
2013.12.25	Zhaoqing	Collapse of port station	6 deaths, 1 injury, direct economic loss of 17 million yuan
2015.12.20	Guangming New District, Shenzhen	Landslide	73 deaths, 4 missing and 17 injured, direct economic loss of 881 million yuan
2019.3.21	Yangzhou, Jiangsu	Scaffold falling in construction site	4 deaths and 7 injuries
2019.3.24	Tieling, Shenyang	Fall from height of a construction site	3 deaths
2019.4.25	Hengshui City, Hebei Province	Fall from height of lift cage	11 deaths, 2 injuries, direct economic loss of about 18 million yuan
2019.5.16	Zhaohua road, Shanghai	Collapse accident of a factory building	12 deaths, 10 serious injuries, 3 slight injuries, million yuan direct economic loss of 34.3
2019.10.28	Guanshan Lake District of Guiyang	Collapse of underground parking lot	8 deaths and 2 injures
2019.6.16	Langfang, Hebei	Slope collapse of foundation pit	3 deaths
2020.5.2	Tianlang City Garden, Xi’an	Elevator Fall from height	3 deaths and 1 serious injury
2020.5.7	Qiemo County, Autonomous Region	Poison of drainage pump station	6 deaths and 1 light injury
2020.5.18	Xinghualing District, Taiyuan City	Collapse of demolition	6 deaths, 7 serious injuries and 12 light injuries
2020.5.19	Dalian Yuda Construction Ltd	Collapse of roof concrete	6 deaths, 3 serious injuries and 15 light injuries
2020.5.21	Xuzhou Municipal Engineering Corporation	Poison of sewage pipes	5 deaths
2021.5.14	Jining, Shandong	Fall from heigh	1 death
2021.6.15	Yushan Road, Nanjing	Foundation Collapse	2 deaths

Although the risk factor analysis in construction accidents is an over-researched problem addressed by updated AI-driven machine learning algorithms in the literature [4, 5, 34], there still remains value to investigate the relationship behind the causing factors due to the variability and complexity of this problem. In addition, to the best of our knowledge there is no formal comparative analysis between the factors causing construction accidents and those causing the severe accidents. In this paper 653 typical safety accidents in construction industry of China are statistically analyzed. The construction safety accidents are fully investigated and formulated with a set of country-specific factors and then used as the source data of association analysis. The main causes contributing to construction safety accidents, particular severe accidents are identified from the 39 common factors using association rules. The beneficial rules can be used in safety inspection to reduce the occurrence of causing factors and improve the accident prevention measures. The contributions of the paper are (1) find out the main factors contributing to accident causation in Chinese construction industry; (2) analyze the importance of causing factors with respect to severe safety accidents; (3) propose some strategies for accident reduction based on the findings.

The rest of this paper is organized as follows. Section 2 outlines the methodology of the study. Section 3 describes the data of construction safety accidents. Section 4 is devoted to some analytic results. Finally, section 5 concludes the paper along with some suggestions and highlights the future directions.

Table 2

Rank of construction safety accidents in China

Rank	Accident	Description	Treatment
1	General accidents	deaths $<$ 3, or serious injuries $<$ 10, or direct economic losses $<$ 10 million yuan	Report to the municipal and handle at county level
2	Major accident	3 $\leqslant$ deaths $<$ 10, or 10 $\leqslant$ serious injuries $<$ 50, or 10 million yuan $\leqslant$ direct economic losses $<$ 50 million yuan	Report to provincial and handle at municipal level
3	Serious accidents	10 $\leqslant$ deaths $<$ 30, or 50 $\leqslant$ serious injuries $<$ 100, or 50 million yuan $\leqslant$ direct economic losses $<$ 100 million yuan	Report to the State Council and handle at the provincial level
4	Very serious accidents	deaths $\geqslant$ 30, or serious injuries $\geqslant$ 100, or direct economic losses $\geqslant$ 100 million yuan	Report to the State Council for handling

2. Association rules

Let $I$ be the set of all items, and $D$ the set of all records. An itemset is a non-empty set of items and an itemset of $K$ items is called k-itemset. Given an itemset $X$ and a record $t$ , $t$ contains $X$ if and only if $X\subseteq T$ . Association rule in the form of $X\rightarrow Y$ , where $X$ is called the left-hand-side (LHS) and $Y$ the right-hand-side (RHS), implies the potential relationship between two itemsets $X$ and $Y$ in $D$ .

Some quantitative criteria including support, confidence and lift are used to define the statistical significance of association rules. Support is defined as the percentage of records in $D$ that contains $X$ . From a statistical point of view, support is equal to the probability of X occurring in D. The support threshold (min_sup) denotes the least importance of itemsets. Given an itemset $X$ and min_sup, if the support of $X$ is no less than min_sup, then $X$ is called frequent itemset. The support of rule $X\rightarrow Y$ indicates the possibility that $X$ and $Y$ occur in the same record, i.e., $\textit{support}(X\cup Y)$ .

The confidence of $X\rightarrow Y$ is defined as the possibility that a record containing $X$ also contains $Y$ , in other words, the conditional probability of $Y$ given $X$ . The confidence threshold min_conf represents the minimum confidence of association rules.

$\textit{Confidence}(X\rightarrow Y)=\frac{\textit{Support}(X\cup Y)}{\textit{% Support}(X)}$

Given a data set $D$ , association analysis intends to find the association rules satisfying the support and confidence no less than the corresponding user-defined thresholds.

The lift of $X\rightarrow Y$ is defined as the ratio of confidence and the support of $Y$ . Lift describes the influence of $X$ on $Y$ . If lift is greater than 1, then $X$ and $Y$ are positively correlated, that is to say, the occurrence of $X$ contributes to the occurrence of $Y$ . If lift is less than 1, $X$ and $Y$ are negatively correlated, in other words, the occurrence of $X$ may lead to the absence of $Y$ . If lift is equal to 1, $X$ and $Y$ are independent. Normally the lift of rules should be more than the lift threshold ( $\textit{min\_lift}>1$ ) indicating the positive correlation between LHS and RHS.

$\textit{Lift}(X\rightarrow Y)=\frac{\textit{Support}(X\cup Y)}{\textit{Support% }(X)\times\textit{Support}(Y)}$

In this study, association rules are used to analyze the relationship among causing factors (items) hidden in the construction accidents (records).

3. Data description

In Chinese Production Safety Law the severity of construction safety accidents is defined as four ranks with respect to the number of deaths and serious injuries, and direct economic losses as shown in Table 2. In this study, the first three ranks (i.e., very serious accidents, serious accidents, and major accidents) are combined to severe accidents due to the minority of data. The data is organized based on the source released in [31] by adding the severe construction safety accidents since 2016 reported in the website of Ministry of Emergence Management and The State Administration of Work Safety. The resulting data set contains 653 accidents, including 221 severe accidents and 432 general accidents. From the perspective of accident type, falling from height, collapse, object strike, and lifting injury are the main accidents in construction industry, accounting for up to 80% of the total number.

Table 3
Property and causing factors of construction safety accidents

	Value	Percentage
Accident severity	General	66.16%
	Severe	33.84%
Accident type	Falling from height	32.5%
	Collapse	26.2%
	Object strike	8.88%
	Lifting injury	11.94%
	Electric shock	6.43%
	Poisoning choke	5.67%
	Vehicle injury	2.45%
	Mechanical injury	1.38%
	Drowning	1.07%
	Explosion	1.68%
	Fire	0.61%
Causing factors	O11 No special safety department	7.81%
	O12 Default of project management team	19.75%
	O21 No enough professional safety engineers	25.42%
	O22 Default of safety engineers	45.18%
	O31 Unqualified workers	26.95%
	O32 Non-compliance of workers	55.13%
	O41 No enough supervision engineers	9.95%
	O42 Inefficient supervision work	29.71%
	T11 Delayed or low-quality drawings	3.37%
	T12 Incomplete construction plan	19.75%
	T13 Inadequate execution of construction plan	37.06%
	T14 Improper installation of temporary facilities	1.38%
	T21 Improper construction procedure arrangement	6.13%
	T22 Inadequate special safety meeting	0.92%
	T23 Inadequate safety inspection	44.41%
	T31 Inadequate hazard recognition	23.58%
	T32 Delayed hazard elimination	35.99%
	R11 Improper operation of mechanical equipment	7.04%
	R12 Delayed maintenance of mechanical equipment	3.98%
	R21 Low-quality construction materials	3.52%
	R22 Improper management of dangerous materials	0.31%
	R31 Inadequate distribution of PPE	9.65%
	R32 Inadequate use of PPE	21.59%
	R33 Inadequate safety warning signs	5.82%
	R34 Inadequate onsite safety protection	36.75%
	R35 Severe onsite work conditions	7.04%
	C11 Schedule pressure	0.92%
	C12 Cost pressure	0.15%
	C13 Inadequate special budget for safety	1.07%
	C21 Unqualified contractor	9.34%
	C22 Multi-level subcontracting	5.05%
	C23 Inadequate subcontractor management	14.40%
	S11 Lacking safety awareness	34.61%
	S12 Lacking safety knowledge	13.94%
	S21 Inadequate safety training	33.08%
	S22 Lacking safety examination	7.96%
	E11 Inadequate accident prevention methods	10.72%
	E12 Inadequate emergency rescue plan	7.81%
	E21 Improper accident reporting and handling	3.37%

As described in [31] the construction safety management system is composed of 6 subsystems, namely Organisation and behaviour (OB), Technical management (TM), Resource support (RS), Contract Management (CM), Safety training (ST), and Emergency management (EM). Each subsystem can be divided into several factors and subfactors. Consequently, each record contains the accident type (Collapse, Object strike, Lifting injury, Electric shock, Poisoning choke, Vehicle injury, Mechanical injury, Drowning, Explosion, Fire), the severity (general, severe), and 39 causing factors. As seen from the percentage values in Table 3, the top four critical causing factors, namely O32 Non-compliance of workers, O22 Default of safety engineers, T23 Inadequate safety inspection, T13 Inadequate execution of construction plan, account for 55.13%, 45.18%, 44.41%, 37.06% respectively in the total construction safety accidents. As outlined in Fig. 2, construction safety accidents caused by four to eight factors account for 70.6% in all accidents. The rough relationship among the factors can be represented by Venn diagram, a common way to show the mathematical or logical relationship between different sets. For instance, Fig. 3 outlines the co-occurrence of the four critical causing factors in accidents. Due to the complexity of causing analysis for construction safety accidents, it is of necessity to explore the hidden relation from the data. In this study, association analysis is used to reveal and quantify the co-occurrence relationship behind the factors from the data.

Figure 2.

Number of causing factors of 653 construction safety accidents.

Figure 3.

Venn diagram of four critical factors.

4. Results and discussions

In this section, association rule mining is utilized to recognize the hidden relationship behind the causing factors of construction safety accidents, and then evaluate the importance of factors with respect to the severity of accidents. Figure 4 depicts the overall methodology, including (1) data collection and pre-processing, (2) cause analysis of construction site accidents to recognize the critical factors and closely related causing factors, (3) significant factors causing severe construction accidents, (4) comparative analysis and safety strategy discussion.

Figure 4.

Flowchart of the proposed methodology.

Table 4

Association rules of construction safety accidents

No.	LHS		RHS	Sup	Conf	Lift	Count
Rule 14	{O31 Unqualified workers}	$\rightarrow$	{O32 Non-compliance of workers}	0.175	0.648	1.175	114
Rule 21	{T23 Inadequate safety inspection, S11 Lacking safety awareness}	$\rightarrow$	{O32 Non-compliance of workers}	0.112	0.619	1.122	73
Rule 23	{S11 Lacking safety awareness}	$\rightarrow$	{O32 Non-compliance of workers}	0.210	0.606	1.100	137
Rule 24	{O12 Default of project management team}	$\rightarrow$	{O32 Non-compliance of workers}	0.118	0.597	1.0827	77
Rule 4	{T23 Inadequate safety inspection, T32 Delayed hazard elimination}	$\rightarrow$	{O22 Default of safety engineers}	0.119	0.605	1.338	78
Rule 6	{T13 Inadequate execution of construction plan, T23 Inadequate safety inspection}	$\rightarrow$	{O22 Default of safety engineers}	0.101	0.574	1.270	66
Rule 8	{R32 Inadequate use of PPE}	$\rightarrow$	{O22 Default of safety engineers}	0.121	0.560	1.240	79
Rule 11	{T23 Inadequate safety inspection}	$\rightarrow$	{O22 Default of safety engineers}	0.242	0.545	1.206	158
Rule 17	{O21 Inadequate professional safety engineers}	$\rightarrow$	{O22 Default of safety engineers}	0.132	0.518	1.147	86
Rule 18	{O42 Inefficient supervision work}	$\rightarrow$	{O22 Default of safety engineers}	0.153	0.515	1.141	100
Rule 19	{O32 Non-compliance of workers, T23 Inadequate safety inspection}	$\rightarrow$	{O22 Default of safety engineers}	0.126	0.513	1.134	82
Rule 22	{T32 Delayed hazard elimination}	$\rightarrow$	{O22 Default of safety engineers}	0.182	0.506	1.120	119
Rule 1	{O22 Default of safety engineers, T32 Delayed hazard elimination}	$\rightarrow$	{T23 Inadequate safety inspection}	0.119	0.655	1.476	78
Rule 5	{O22 Default of safety engineers, T13 Inadequate execution of construction plan}	$\rightarrow$	{T23 Inadequate safety inspection}	0.101	0.574	1.292	66
Rule 7	{R32 Inadequate use of PPE}	$\rightarrow$	{T23 Inadequate safety inspection}	0.119	0.553	1.246	78
Rule 9	{T32 Delayed hazard elimination}	$\rightarrow$	{T23 Inadequate safety inspection}	0.198	0.549	1.236	129
Rule 10	{O22 Default of safety engineers}	$\rightarrow$	{T23 Inadequate safety inspection}	0.242	0.536	1.206	158
Rule 12	{O32 Non-compliance of workers, S11 Lacking safety awareness}	$\rightarrow$	{T23 Inadequate safety inspection}	0.112	0.533	1.200	73
Rule 13	{S11 Lacking safety awareness}	$\rightarrow$	{T23 Inadequate safety inspection}	0.181	0.522	1.176	118
Rule 15	{O12 Default of project management team}	$\rightarrow$	{T23 Inadequate safety inspection}	0.103	0.519	1.170	67
Rule 16	{O32 Non-compliance of workers, T32 Delayed hazard elimination}	$\rightarrow$	{T23 Inadequate safety inspection}	0.103	0.511	1.153	67
Rule 20	{O22 Default of safety engineers, O32 Non-compliance of workers}	$\rightarrow$	{T23 Inadequate safety inspection}	0.126	0.500	1.126	82
Rule 2	{T12 Incomplete construction plan}	$\rightarrow$	{T13 Inadequate execution of construction plan}	0.107	0.543	1.464	70
Rule 3	{S21 Inadequate safety training}	$\rightarrow$	{T13 Inadequate execution of construction plan}	0.167	0.505	1.362	109

Figure 5.

Grouped matrix of 37 association rules.

4.1 Cause analysis of construction safety accidents

The Apriori algorithm implemented in arulesViz package [8] is used to find association rules from construction safety accidents. By setting min_sup as 0.1 and min_conf as 0.5, a total of 37 rules are found implying the relationship among the causing factors. Figure 5 shows the grouped matrix of rules which clusters the itemsets shown in the $x$ -axis (LHS) and the $y$ -axis (RHS) respectively. The vertices in the matrix represent the association rules composed of the corresponding itemsets in LHS and RHS. The size and color of vertices denote the support and lift of association rules respectively. It is noted that the discovered 37 rules take the four critical causing factors, namely O22 Default of safety engineers, O32 Non-compliance of workers, T13 Inadequate execution of construction plan, T23 Inadequate safety inspection, as RHS of association rules. In order ro provide a clear representation and explanation of rules, set min_lift $=$ 1.1 resulting in 24 rules listed in Table 4. With the network visualization with respect to each critical factor shown in Figs 6–9 respectively, the association rules can provide more understandable results for investigating the causes behind construction safety accidents.

In general, The found association rules are beneficial to improving the efficiency of safety inspection and optimizing the safety early warning system. Some evidences can be found from the results. (1) T12 Incomplete construction plan and S21 Inadequate safety training have positive impact on T13 Inadequate execution of construction plan with the lift up to 1.464 (Rule 2) and 1.362 (Rule 3) respectively, that is to say, the incomplete construction plan or the inadequate safety training may lead to the inadequate execution of construction plan. (2) The causing factors O31 Unqualified workers, O12 Default of project management team, T23 Inadequate safety inspection, and S11 Lacking safety awareness will probably lead to O32 Non-compliance of workers (Rule 14, Rule 21, Rule 23, and Rule 24). For example, if the workers are unqualified, then they are probably non-compliance with the confidence $=$ 64.8% and lift $=$ 1.175 (Rule 14). (3) Many causing factors involved in 10 rules have positive impact on the occurrence of T23 Inadequate safety inspection. Particularly, if both O22 Default of safety engineers and T32 Delayed hazard elimination exist, then T23 Inadequate safety inspection will probably occur with confidence $=$ 65.5% and lift $=$ 1.476 (Rule 1). Additionally, T13 Inadequate execution of construction plan (Rule 5), R32 Inadequate use of PPE (Rule 7), T32 Delayed hazard elimination (Rule 9), S11 Lacking safety awareness (Rule 12, Rule 13), O12 Default of project management team (Rule 15) and O32 Non-compliance of workers (Rule 16, Rule 20) have impact on T23 Inadequate safety inspection. (4) Regarding O22 Default of safety engineers, the most closely related factor is T23 Inadequate safety inspection, occurring in four association rules (Rule 4, Rule 6, Rule 11, and Rule 19) among the totally 8 rules. (5) Figure 10 shows the network visualization of the total 24 association rules with respect to the four critical factors, where vertices represents itemsets, directed edges represent rules between itemsets with size and color indicating support and lift of rules respectively. It is disclosed that the four critical causing factors are closely related, occurring in both LHS and RHS of association rules (shown in Table 5). It implies that eliminating one will reduce the probability of the others, so as to consequently prevent the occurrence of construction safety accidents.

Figure 6.

Network visualization: O22 default of safety engineer.

Figure 7.

Network visualization: O32 non-compliance of workers.

Figure 8.

Network visualization: T13 inadequate execution of construction plan.

Figure 9.

Network visualization: T23 inadequate safety inspection.

Figure 10.

Network visualization of 24 association rules.

Table 5

Impacting factors of four critical factors

Critical factors	Impacting factors
(RHS)	(LHS)
[l]O32 Non-complianceof workers	O31 Unqualified workers
[l]O32 Non-complianceof workers	T23 Inadequate safety inspection
	S11 Lacking safety awareness
	O12 Default of project management team
[l]O22 Default of safetyengineers	T23 Inadequate safety inspection
[l]O22 Default of safetyengineers	T32 Delayed hazard elimination
	T13 Inadequate execution of construction plan
	R32 Inadequate use of PPE
	O21 No enough professional safety engineers
	O42 Inefficient supervision work
	O32 Non-compliance of workers
[l]T23 Inadequate safetyinspection	O22 Default of safety engineers
[l]T23 Inadequate safetyinspection	T32 Delayed hazard elimination
	T13 Inadequate execution of construction plan
	R32 Inadequate use of PPE
	O32 Non-compliance of workers
	S11 Lacking safety awareness
	O12 Default of project management team
[l]T13 Inadequate execution of construction plan	T12 Incomplete construction plan
[l]T13 Inadequate execution of construction plan	S21 Inadequate safety training

Table 6

Causing factors wrt. severe construction safety accidents

LHS		RHS	Sup	Conf	Lift
R22 Inadequate management for dangerous materials	$\rightarrow$	Severe	0.00	1.00	2.95
T11 Delayed or low-quality drawings	$\rightarrow$	Severe	0.03	0.86	2.55
O41 Inadequate supervision engineers	$\rightarrow$	Severe	0.09	0.86	2.55
S22 Lacking safety examination	$\rightarrow$	Severe	0.06	0.77	2.27
C21 Unqualified contractor	$\rightarrow$	Severe	0.07	0.75	2.23
E21 Improper accident reporting and handling	$\rightarrow$	Severe	0.02	0.73	2.15
C22 Multi-level subcontracting	$\rightarrow$	Severe	0.03	0.67	1.97
E12 Inadequate emergency rescue plan	$\rightarrow$	Severe	0.05	0.63	1.85
T12 Incomplete construction plan	$\rightarrow$	Severe	0.11	0.57	1.67
T14 Improper installation of temporary facilities	$\rightarrow$	Severe	0.01	0.56	1.64
R11 Improper operation of mechanical equipment	$\rightarrow$	Severe	0.04	0.54	1.61
C11 Too tight schedule	$\rightarrow$	Severe	0.00	0.50	1.48
E11 Inadequate accident prevention measurements	$\rightarrow$	Severe	0.05	0.50	1.48
R35 Severe onsite work conditions	$\rightarrow$	Severe	0.03	0.46	1.35
O42 Inefficient supervision work	$\rightarrow$	Severe	0.13	0.45	1.33
T31 Inadequate hazard recognition	$\rightarrow$	Severe	0.10	0.44	1.30
O11 No special safety department	$\rightarrow$	Severe	0.03	0.43	1.27
C13 Inadequate special fee for safety	$\rightarrow$	Severe	0.00	0.43	1.27
R12 Delayed maintenance for mechanical equipment	$\rightarrow$	Severe	0.02	0.42	1.25
T21 Improper construction procedure arrangement	$\rightarrow$	Severe	0.02	0.40	1.18
O31 Unqualified workers	$\rightarrow$	Severe	0.11	0.40	1.18
R31 Inadequate distribution of PPE	$\rightarrow$	Severe	0.04	0.40	1.17
C23 Inadequate subcontractor management	$\rightarrow$	Severe	0.06	0.39	1.16
R21 Low-quality construction materials	$\rightarrow$	Severe	0.01	0.39	1.16
S21 Inadequate safety training	$\rightarrow$	Severe	0.13	0.38	1.12
T13 Inadequate execution of construction plan	$\rightarrow$	Severe	0.13	0.36	1.07
O12 Default of project management team	$\rightarrow$	Severe	0.07	0.36	1.05
O32 Non-compliance of workers	$\rightarrow$	Severe	0.18	0.32	0.95
S12 Lacking safety knowledge	$\rightarrow$	Severe	0.04	0.32	0.94
R34 Inadequate onsite safety protection	$\rightarrow$	Severe	0.12	0.31	0.91
O21 Inadequate professional safety engineers	$\rightarrow$	Severe	0.07	0.29	0.85
T23 Inadequate safety inspection	$\rightarrow$	Severe	0.12	0.27	0.80
T32 Delayed hazard elimination	$\rightarrow$	Severe	0.10	0.27	0.79
R33 Inadequate safety warning signs	$\rightarrow$	Severe	0.01	0.24	0.70
R32 Inadequate use of PPE	$\rightarrow$	Severe	0.05	0.22	0.65
S11 Lacking safety awareness	$\rightarrow$	Severe	0.07	0.20	0.60
O22 Default of safety engineers	$\rightarrow$	Severe	0.09	0.20	0.60
T22 Inadequate special safety meeting	$\rightarrow$	Severe	0.00	0.17	0.49
C12 Too low contract price	$\rightarrow$	Severe	0.00	0.00	0.00

4.2 Causing factors with respect to accident severity

In Chinese construction industry, the severity of accidents has particular importance on the safety production license and qualification of enterprises. According to the regulations on safety production management of construction projects issued by the Chinese State Council, in case of a major (or more serious) safety accident the enterprise shall temporarily suspend the work safety license for 60 to 90 days, or even lose the work safety license and bidding qualification in the construction market. In consequence, the sever accidents should be given more attention than general accidents and controlled strictly. For this purpose, the causing factors impacted mostly on the severity of accidents are analyzed using association rules and measured by lift.

Table 6 shows the support, confidence, and lift of rules taking each causing factor as LHS and severe accident as RHS. The rules are listed in descending order of lift, which implies the positive relevance with severe accident with the value more than 1, and negative relevance with the value less than $1$ . Setting min_sup as 0.2, min_conf $=$ 0.5, and min_lift as 1.5, a total of 9 causing factors can be considered as the main factors contributing mostly to the severity of construction safety accidents, which are T11 Delayed or low-quality drawings, T12 Incomplete construction plan in Technical management subsystem, O41 Inadequate supervision engineers in Organisation and behaviour subsystem, S22 Lacking safety examination in Safety training subsystem, E12 Inadequate emergency rescue plan and E21 Improper accident reporting and handling in Emergency management subsystem, R11 Improper operation of mechanical equipment in Resource support subsystem, C21 Unqualified contractor and C22 Multi-level subcontracting in Contract Management subsystem.

The quality of construction drawing design is the foundation of construction project. The design drawings must comply with the regulations on the depth of construction engineering design documentation. However, due to the short design cycle and heavy task at present the problem of low-quality drawing is common in construction industry. It is a major issue to improve the responsibility of designers, strengthen the proofreading of drawings, and reduce the mistakes.

The supervision engineers play an important role in the process of construction and ensure the overall quality of the construction project. However, the construction supervision in China is suffered from inadequate or unqualified supervision engineers in order to reduce the investment. To properly complete the project supervision, the quality supervision is worthy of attention in the whole life cycle of construction project [25].

Multi-tier subcontracting system is widely adopted within the construction industry in the world for the better efficiency due to the specialization of subcontractors [14]. However it usually makes difficult to establish effective safety supervision system so that the safety accidents caused by subcontracting are common. Therefore, it is necessary for general contractor to control safety accidents under the subcontracting system.

It should be noted that the emergency rescue plan and proper accident reporting and handling are essential to the severity of construction accidents, not only effectively preventing the occurrence of construction safety accidents but also minimizing the casualties and property losses once the accidents occur.

For comparison Table 7 lists the top 10 causing factors related to the severity of construction accidents using the proposed Association Rule $+$ Lift (M1) method and five attribute selection methods respectively, namely, CfsSubsetEval $+$ BestFirst (M2), Ranker $+$ InfoGainEval (M3), Ranker $+$ ReliefEval (M4), Ranker $+$ Correlation (M5), Ranker $+$ GainRatioEval (M6). Among these factors, T11, O41, S22, C21, E21, C22, E12, T12 are considered to have significant impact on the severity of accidents by the majority of the methods. Some factors such as S11, O22, C12 are considered as important only by attribute selection methods. This is due to the different measurement adopted, i.e., the positive impact on the sever accidents (M1), and the impact on the consequences of accidents including both sever accidents and general accidents (M2–M6).

Table 7
Importance of factors to the severity of construction accidents using six methods (M1: Association rule $+$ lift, M2: CfsSubsetEval $+$ bestFirst, M3: Ranker $+$ InfoGainEval, M4: Ranker $+$ ReliefEval, M5: Ranker $+$ Correlation, M6: Ranker $+$ GainRatioEval)

Rank	M1	M2	M3	M4	M5	M6
1	T11	O22	O41	O22	O41	O41
1	O41	O41	C21	R32	C21	R22
3	S22	T11	O22	S11	S22	T11
4	C21	T12	S22	T23	O22	S22
5	E21	T32	T12	O41	T12	C21
6	C22	R11	S11	C21	T11	E21
7	E12	C21	T11	T32	S11	C22
8	T12	C22	E12	O32	E12	C12
9	R11	S11	C22	T12	C22	T12
10	C11	S22	O42	O42	E21	E12

5. Conclusions

The number of construction safety accidents occurred in China is increasing significantly in recent years, therefore, it is of utmost importance to implement corrective and preventive measures in order to prevent the accidents in the future. This research intends to reveal unknown relationships among 39 common causing factors contributing to construction safety accidents by analyzing the data of 653 past accidents from 2008 to 2020 in China. By utilizing arulesViz package as the data mining software tool, the preferential relation are found behind the construction casualty causes.

Our findings yields some important insights to construction safety management and decision support. In general, the organizing and behaviour, particularly non-compliance of workers and default of safety engineer, and technical management, particularly inadequate safety inspection and inadequate execution of construction plan, are the critical factors causing safety accidents in construction industry. Besides, 9 factors are found closely related to the critical factors, including unqualified workers, default of project management team, inadequate safety inspection, lacking safety awareness, inadequate use of PPE, delayed hazard elimination, no enough professional safety engineers, inefficient supervision work, and incomplete construction plan. From the perspective of accident severity, the most impacting factors are delayed or low-quality drawings, incomplete construction plan, inadequate supervision engineers, lacking safety examination, inadequate emergency rescue plan, improper accident reporting and handling, improper operation of mechanical equipment, unqualified contractor and multi-level subcontracting.

Based on the findings of the study, some strategies can be put forward. Firstly, the results shows that human unsafe behavior acts as a main factor on construction accidents, referring to improper operation of workers, command error of engineers, poor quality of workers, etc. In China there is a large proportion of migrant workers in construction industry. According to the report of the National Bureau of statistics, the total number of migrant workers in 2021 is 273.95 million, of which 18.3% are in the construction industry. Migrant workers generally have low level of education and low proportion of vocational training resulting in inadequate awareness and knowledge of safety. The construction enterprises should pay attention to the non-compliance of workers by strengthening the safety training of employee, improving their safety awareness and professional skills, and preventing the occurrence of unsafe behaviors. The safety engineers who are responsible for safety plan design, safety inspection, and accident elimination play an important role in the process of construction project. The default of safety engineers can be largely avoided by adequate safety inspection and supervision, proper utilization of PPE, and elimination of potential risks.

Secondly, the importance of PPE is verified in the study as was widely observed [28]. The statistics illustrate that inadequate usage of PPE accounts for around 35% of the total accidents in Chinese construction industry. The inadequate usage of PPE including unqualified PPE, incorrect wearing of PPE, delayed distribution of PPE is another factor which endangers the safety and health of workers. The enterprises and supervisor department should highlights PPE’s importance in construction projects so as to reduce the accidents and occupational injuries of on-site workers.

Thirdly, the enterprises should improve the safety management system, formulate reasonable construction organization design and construction scheme, implement safety risk management, identify and confirm the hazard factors in the construction process. In order to diminish the severe safety accidents, sufficient attention should be given on improving the low-quality drawings, completing the construction and emergency rescue plan, completing safety examination with adequate supervision engineers, avoiding improper operation of mechanical equipment, unqualified and multi-level contractor.

This study concentrates on the relationship among causing factors contributing to the common construction safety accidents, particular the severe accidents. In the next study the relationship between the causing factors and the type of accidents will be investigated. Additionally the discovered association rules will be integrated in safety inspection system for risk prediction.

Footnotes

Acknowledgements

This work was supported by National Office of Philosophy and Social Sciences (19AZD019), National Ethnic Affairs Commission (2020-GMB-015).

References

Agrawal

Srikant

. Fast algorithms for mining association rules. 20th Very Large Data Bases Conferences. 1994; 487-499.

Alomari

Gambatese

Nnaji

Tymvios

. Impact of risk factors on construction worker safety: A delphi rating study based on field worker perspective. Arabian Journal for Science and Engineering. 2020; 45.

Ammad

Alaloul

Saad

Qureshi

. Personal protective equipment (PPE) usage in construction projects: A scientometric approach. Journal of Building Engineering. 2021; 35: 102086.

Changhai

Shenping

. Factors correlation mining on maritime accidents database using association rule learning algorithm. Cluster Computing. 2019; 22: 4551-4559.

Cheng

Kusoemo

Gosno

. Text miningbased construction site accident classification using hybrid supervised machine learning. Automation in Construction. 2015; 118.

Dunham

Gruenwald

Hossain

. A survey of association rules. Descriptive Data Mining. 2019; 161-167.

Fang

Xie

Huang

. Factor analysis-based studies on construction workplace safety management in china. International Journal of Project Management. 2004; 22(1): 43-49.

Hahsler

Karpienko

. Visualizing association rules in hierarchical groups. Journal of Business Economics. 2017; 87(3): 317-335.

Hallowell

Hinze

Baud

Wehle

. Proactive construction safety control: Measuring, monitoring, and responding to safety leading indicators. Journal of Construction Engineering and Management. 2013; 139(10): 04013010.

10.

Heinrich

. Industrial accident prevention: A scientific approach. McGraw-Hill. 1931.

11.

Huang

Perng

. Analysis of association rule in fire accidents-evidence from taipei city. Advanced Materials Research. 2014; 1030-1032: 2407-2410.

12.

Kaur

Madan

. Association rule mining: A survey. International Journal of Hybrid Information Technology. 2015; 6(7): 239-242.

13.

Leveson

Dulac

Marais

Carroll

. Moving beyond normal accidents and high rrliability organizations: A systems approach to safety in complex systems. Organization Studies. 2009; 30(2-3): 227-249.

14.

Lew

Lai

Toh

, et al. Quality performance of multi-layered subcontracting practices in malaysian construction industry. IOP Conference Series Earth and Environmental Science. 2020; 498: 012092.

15.

Manu

Ankrah

Proverbs

, et al. An approach for determining the extent of contribution of construction project features to accident causation. Safety Science. 2010; 48(6): 687-692.

16.

Marzouk

. Simulation of labor evacuation: The case of housing construction projects. HBRC Journal. 2018; 14(2): 198-206.

17.

Mirabadi

Sharifian

. Application of association rules in iranian railways (RAI) accident data analysis. Safety Science. 2010; 48(10): 1427-1435.

18.

Mitropoulos

Abdelhamid

Howell

. Systems model of construction accident causation. Journal of Construction Engineering and Management. 2005; 131(7): 816-825.

19.

Molenaar

Park

Washington

. Framework for measuring corporate safety culture and its impact on construction safety performance. Journal of Construction Engineering and Management. 2009; 135(6): 488-496.

20.

Perlman

Sacks

Barak

. Hazard recognition and risk perception in construction. Safety Science. 2014; 64: 13-21.

21.

Qiao

Liu

Lia

, et al. Using data mining techniques to analyze the influencing factor of unsafe behaviors in chinese underground coal mines. Resources Policy. 2021; 210-216.

22.

Sertyesilisik

Tunstall

McLouglin

. An investigation of lifting operations on UK construction sites. Safety Science. 2010; 48(1): 72-79.

23.

Shin

Park

Seo

, et al. Association rules mined from construction accident data. KSCE Journal of Civil Engineering. 2017; 22(1): 1-13.

24.

Wang

Zeng

Xiong

, et al. Finding main causes of elevator accidents via multi-dimensional association rule in edge computing environment. China Communications. 2017; 11: 45-53.

25.

Wei

Sun

Wei

, et al. The quality supervision model construction of construction project in the whole life cycle. Journal of Physics Conference Series. 2020; 1549: 032023.

26.

Weng

Zhu

Yan

, et al. Investigation of work zone crash casualty patterns using association rules. Accident Analysis and Prevention. 2016; 92: 43-52.

27.

Yaacob

Rusli

Bohari

. A review analysis of accident factor on road accident cases using haddon matrix approach. 2nd International Conference on the Future of ASEAN (ICoFA). Springer Singapore. 2018; 2: 55-65.

28.

Yang

Shirowzhan

Sepasgozar

BHC

. Automated PPE-tool pair check system for construction safety using smart ioT. Journal of Building Engineering. 2020; 32(3): 101721.

29.

Zahoor

Chan

Ran

, et al. The factors contributing to construction accidents in pakistan: Their prioritization using the delphi technique. Engineering Construction & Architectural Management. 2017; 24(3): 463-485.

30.

Zhang

Peixui

, et al. Applicability of accident analysis methods to chinese construction accidents. Journal of Safety Research. 2019; 68: 187-196.

31.

Zhang

Zhu

Zhang

Zhao

. Identification of critical causes of construction accidents in China using a model based on system thinking and case analysis. Safety Scienc. 2020; 121: 606-618.

32.

Zhang

Liu

. Analysis of multi-dimensional association rule in marine casualties. International Conference on Transportation Information & Safety. 2011; 2697-2705.

33.

Guo

Zhou

Irizarry

. A network-based approach to modeling safety accidents and causations within the context of subway construction project management. Safety Science. 2021; 139(10): 105261.

34.

Zhu

Hou

. Application of machine learning techniques for predicting the consequences of construction accidents in china. Process Safety and Environmental Protection. 2021; 145: 293-302.

Rank	M1	M2	M3	M4	M5	M6
1	T11	O22	O41	O22	O41	O41
1	O41	O41	C21	R32	C21	R22
3	S22	T11	O22	S11	S22	T11
4	C21	T12	S22	T23	O22	S22
5	E21	T32	T12	O41	T12	C21
6	C22	R11	S11	C21	T11	E21
7	E12	C21	T11	T32	S11	C22
8	T12	C22	E12	O32	E12	C12
9	R11	S11	C22	T12	C22	T12
10	C11	S22	O42	O42	E21	E12

Rank	M1	M2	M3	M4	M5	M6
1	T11	O22	O41	O22	O41	O41
1	O41	O41	C21	R32	C21	R22
3	S22	T11	O22	S11	S22	T11
4	C21	T12	S22	T23	O22	S22
5	E21	T32	T12	O41	T12	C21
6	C22	R11	S11	C21	T11	E21
7	E12	C21	T11	T32	S11	C22
8	T12	C22	E12	O32	E12	C12
9	R11	S11	C22	T12	C22	T12
10	C11	S22	O42	O42	E21	E12

Cause analysis of construction safety accidents in China using association rules

Abstract

Keywords

1. Introduction

3. Data description

Table 3 Property and causing factors of construction safety accidents

Table 7 Importance of factors to the severity of construction accidents using six methods (M1: Association rule + lift, M2: CfsSubsetEval + bestFirst, M3: Ranker + InfoGainEval, M4: Ranker + ReliefEval, M5: Ranker + Correlation, M6: Ranker + GainRatioEval)

Footnotes

Acknowledgements

References

Table 3
Property and causing factors of construction safety accidents

Table 7
Importance of factors to the severity of construction accidents using six methods (M1: Association rule $+$ lift, M2: CfsSubsetEval $+$ bestFirst, M3: Ranker $+$ InfoGainEval, M4: Ranker $+$ ReliefEval, M5: Ranker $+$ Correlation, M6: Ranker $+$ GainRatioEval)

Rank	M1	M2	M3	M4	M5	M6
1	T11	O22	O41	O22	O41	O41
1	O41	O41	C21	R32	C21	R22
3	S22	T11	O22	S11	S22	T11
4	C21	T12	S22	T23	O22	S22
5	E21	T32	T12	O41	T12	C21
6	C22	R11	S11	C21	T11	E21
7	E12	C21	T11	T32	S11	C22
8	T12	C22	E12	O32	E12	C12
9	R11	S11	C22	T12	C22	T12
10	C11	S22	O42	O42	E21	E12