A Multi-Dimensional Construction Safety Risk Optimization Model

Abstract

Occupational accidents in the construction sector are a significant concern for government agencies and enterprises globally. A detailed assessment of the potential consequences of accidents is essential for supervisory teams. This study presents a novel multidimensional safety assessment optimization model that assesses the cost–benefit relationship of safety measures, considering their impact on workers, company reputation, project cost, project duration, society, and the environment. First, safety risks and measures for primary work items in a typical building construction were determined. The experts subsequently assessed these risks based on precautions. Then, an optimization model was designed using a genetic algorithm and implemented for the risk assessment of a building construction project to identify the optimal measures for reducing risk scores and precautionary costs. Despite the total risk score achieved using the developed approach increased by 17.86% compared to the traditional risk assessment technique, the precautionary measures cost was reduced by 43.60%. Comparing the proposed model with the traditional risk assessment approach, it is observed that the model provides near-optimal risk scores and precautionary costs. The study offers significant implications for both practice and theory by examining risks from multiple perspectives and providing flexibility to users.

1. Introduction

The construction industry contains numerous hazards caused by dangerous working conditions and the use of various tools, equipment, and machinery [1,2,3]. A precise assessment of potential risks is an essential step in accident prevention. There are two basic approaches to risk assessment. The first one is the proactive approach, which aims to eliminate or reduce these risks to an acceptable level by anticipating the risks in the system before undesirable events occur. The second is the reactive approach, which is developed to examine the causes of undesirable events such as work accidents or occupational diseases after they occur and to offer suggestions to prevent their recurrence [4]. To enhance occupational safety performance and prevent potential incidents at construction sites, a proactive strategy must be employed [5,6,7,8,9]. Comprehensive risk assessments, driven by expert insights, can foster a proactive approach to accident avoidance in the construction industry [7,10].

The assessment of occupational risks is considered a complex multi-criteria decision-making problem because it involves multiple risk criteria [11]. Therefore, researchers have undertaken numerous studies to develop a multidimensional risk assessment model designed for the construction sector [12,13,14,15,16,17]. Multidimensional risk assessment models are usually developed by combining conventional methodologies with Multi-Criteria Decision-Making Methods (MCDM) such as Technique for Order Preference Similarity to the Ideal Solution (TOPSIS), Analytic Hierarchy Process (AHP), or Fuzzy Logic [16,18,19]. The dimensions mostly addressed in the models are frequency, severity, detectability, sensitivity to personal protective equipment, and cost. The severity parameter inherently prioritizes the well-being and safety of the workers, and the rating is determined from this perspective. Undoubtedly, the well-being and protection of workers are of the greatest significance and ought to be the primary objective of all efforts related to risk assessment. Notwithstanding many variables in the multidimensional risk assessment literature, some important severity parameters have been neglected. Thus, an occupational accident at the construction site not only threatens the workers’ physical integrity but also can negatively affect the environment, society, company reputation, project budget, and duration [20,21]. For example, an explosion, fire, or chemical spill at a construction site not only affects the health of workers but also has the potential to threaten the air, rivers, and soil in the area. The workers’ families face sociological and economic consequences, and the company’s reputation may suffer significant damage. Moreover, financial losses, compensation, and delays directly affect the project budget and duration, which are key components of project performance. To give another example, equipment damage that does not have a significant impact on workers and is classified as an “acceptable risk” can cause a significant increase in project duration and cost. Therefore, diversifying the severity parameter will allow for a more comprehensive approach to risk assessment.

Another main goal in safety risk assessment is to achieve the greatest benefit at the optimum cost of precaution [22,23]. The expenses of measures and their effectiveness in risk mitigation are crucial for assessing cost–benefit equilibrium. A greater budget allocation to measures results in an increased total project budget. However, increasing the budget does not ensure a proportional reduction in total risk. Achieving a cost–benefit equilibrium can be exceedingly challenging, particularly in complex construction projects where risks vary based on project characteristics and location, necessitating the management of numerous risks. This is an optimization problem that necessitates lowering costs while ensuring that safety is not compromised.

Related literature shows that optimization techniques such as genetic algorithms (GAs) and Particle Swarm Optimization (PSO) have been successfully implemented for such multi-criteria problems to analyze, classify, and find the possible best solution [23,24,25]. Moreover, safety management-related research underlines the fact that by optimizing risk assessment, it is possible to significantly minimize uncertainty arising from several sources, including working conditions, human factors, technological equipment failures, and the environment [15]. Previous research mostly determined the risk score based on traditional dimensions; however, it is crucial to consider additional dimensions as risks affect the project, the organization, and the community [19,26]. Therefore, this study adopted a multi-dimensional severity parameter by considering the impact of risks on the environment, society, company reputation, project cost, and duration. For this purpose, an optimization model based on a multi-objective GA is developed and applied to the risk assessment of a 21-story building construction project. While developing the model, risk assessment documents and literature were considered to identify risks to be encountered, prepare the related questionnaires to collect expert opinions, and base the conceptual framework of the optimization model. This approach ensures that risk assessment techniques consider both the material and moral consequences of accidents.

The remainder of this paper is organized as follows: Section 2 summarizes previous studies on risk assessments in the construction industry, the differences between the present study and previous studies, and its contribution to the literature. Section 3 provides information on the methodology adopted in the study and the data used. Section 4 presents the findings and the discussion based on the evaluation of the model. Ultimately, Section 5, Section 6 and Section 7 conclude by discussing the potential implications, limitations, and future potential of the study.

2. Safety Risk Assessment Practices in the Construction Industry

The identification and assessment of occupational health and safety addresses play an essential role in the construction sector due to its complex structure and hazardous working conditions. Various risk assessment methodologies have been developed to manage risks. These methodologies are classified into two categories: qualitative and quantitative approaches. Efforts to enhance safety risk assessment methodologies in the construction sector indicate that both approaches have substantially contributed [14,17,27,28,29,30]. For example, Ahmed et al. [27] developed a comprehensive risk assessment model combining the Fuzzy Analytic Hierarchy Process (FAHP) and Fuzzy Inference System (FIS) to evaluate different factors. On the other hand, Durmaz and Gölcük [17] introduced a mathematical methodology to prioritize and categorize risk factors by combining the entropy weighting technique and the TOPSIS-Sort B method. Gürcanlı et al. [28] employed an L-type matrix and Fine–Kinney techniques to assess risks in reinforced concrete housing projects. Zermane et al. [30] proposed a two-stage assessment approach incorporating statistical analysis and Fault Tree Analysis (FTA) in risk assessment methodologies. Researchers in this field have sought to enhance the efficacy of risk assessment through the development of novel risk criteria or methodologies. They also developed new risk parameters to improve the completeness of risk assessment and used multi-criteria methodologies for more effective risk prioritization.

2.1. Safety Risk Parameters

Traditional risk assessment approaches determine probability and severity as risk parameters. These are the basic risk parameters and have been extensively utilized across various sectors, including the construction industry [31,32,33,34]. Practitioners, considering probability and severity parameters to be inadequate for risk assessment, include the exposure parameter in their evaluation utilizing the Fine–Kinney method [14,35,36], a widely used quantitative approach [14,17,29,30,31,37,38,39,40]. The exposure parameter quantifies the frequency of worker exposure to hazards. Thus, the probability is supported by statistical data. In another traditional methodology, the Failure Mode and Effects Analysis (FMEA) technique [29,41] incorporates the “detection” parameter into the variables, factoring in the probability of identifying the hazard before its occurrence. Numerous studies have been conducted to enhance the traditional risk parameters, and several alternative risk parameters have been introduced to the construction safety literature (Table 1).

Table 1. Safety risk parameters in construction safety literature.

Risk Parameter	Reference
Probability	[14,16,17,18,19,29,31,35,36,38,40,41,42,43,44]
Severity	[14,16,17,18,19,29,31,35,36,38,40,41,42,43,44]
Frequency/Exposure	[14,16,19,35,36,41,43]
Detection	[16,17,19,29,38,41,43]
Expenses	[42]
Safety barriers	[18]
Safety climate	[18]
Accident percentage	[42]
Safety level	[42]
Hazard criticality	[31]
Sensitivity to the failure of barriers	[16]
Sensitivity to the poor safety climate	[16]
Sensitivity to poor site safety conditions	[16]
Sensitivity to bad environmental conditions	[16]
Worsening factor	[19]
Sensitivity	[44]

Table 1. Safety risk parameters in construction safety literature.

Risk Parameter	Reference
Probability	[14,16,17,18,19,29,31,35,36,38,40,41,42,43,44]
Severity	[14,16,17,18,19,29,31,35,36,38,40,41,42,43,44]
Frequency/Exposure	[14,16,19,35,36,41,43]
Detection	[16,17,19,29,38,41,43]
Expenses	[42]
Safety barriers	[18]
Safety climate	[18]
Accident percentage	[42]
Safety level	[42]
Hazard criticality	[31]
Sensitivity to the failure of barriers	[16]
Sensitivity to the poor safety climate	[16]
Sensitivity to poor site safety conditions	[16]
Sensitivity to bad environmental conditions	[16]
Worsening factor	[19]
Sensitivity	[44]

Pinto [18] integrated the safety barrier and climate with the traditional probability and severity risk parameters. In this way, the author provides a risk assessment that includes the effectiveness of the measures and the safety climate on the construction site. Debnath et al. [42] developed a risk assessment model using the parameters of expenses, accident percentage, and safety level. Similar to the exposure parameter, the accident percentage was determined and integrated into the assessment. The safety level indicates the degree of safety achieved by the measures. The expenses parameter represents costs related to safety equipment, training, maintenance, etc. Sanni-Anibire et al. [31] introduced the hazard criticality parameter, in which the presence of accident causes is determined. The probability parameter is enhanced by examining the existence of accident causes. Mohandes and Zhang [16] noted the sensitivity of hazards to several factors, including barrier failures, poor safety climate, inadequate site safety conditions, and bad environmental conditions. The authors emphasized the necessity of including sensitivity in risk assessment. In another approach akin to Mohandes and Zhang [16], Yao et al. [44] employed a sensitivity parameter that signifies the degree to which interventions related to the causal factors of accidents might affect both the probability of accidents and their consequences. Finally, Mohandes et al. [19] employed the worsening factors proposed by Ouédraogo et al. [45] to assess construction safety risks. Worsening factors refer to hazardous conditions that may exacerbate the immediate consequences of an incident.

Most risk parameters offered in the literature enhance the precision of probability assessment. However, as previously highlighted, the safety risks can also impact the project duration and cost, society, the company, and the environment. Table 2 summarizes the main factors influenced by risks in the construction industry.

Table 2. Main factors influenced by safety risks.

Reference	Project Cost	Project Duration	Company Reputation	Environment	Society
[46]	✓
[47]				✓
[48]	✓	✓
[20]	✓	✓	✓		✓
[26]	✓	✓	✓		✓
[28]	✓	✓
[36]		✓
[49]	✓	✓
[38]	✓		✓	✓	✓
[39]	✓		✓
[50]	✓	✓		✓
[51]			✓		✓
[16]				✓
[19]	✓	✓	✓		✓
[21]	✓	✓	✓		✓
[15]				✓	✓
[52]				✓
[53]	✓	✓
[54]	✓	✓		✓

Table 2. Main factors influenced by safety risks.

Reference	Project Cost	Project Duration	Company Reputation	Environment	Society
[46]	✓
[47]				✓
[48]	✓	✓
[20]	✓	✓	✓		✓
[26]	✓	✓	✓		✓
[28]	✓	✓
[36]		✓
[49]	✓	✓
[38]	✓		✓	✓	✓
[39]	✓		✓
[50]	✓	✓		✓
[51]			✓		✓
[16]				✓
[19]	✓	✓	✓		✓
[21]	✓	✓	✓		✓
[15]				✓	✓
[52]				✓
[53]	✓	✓
[54]	✓	✓		✓

Safety risks on construction sites also have a wide impact. First, they have the potential to affect the project budget and duration significantly. When risk materializes, construction site operations can be interrupted for many days to allow legal investigations, restore the necessary workforce, or replace damaged machinery and equipment. Moreover, compensation paid to workers, machinery and equipment replacement costs, and contractual obligations as a result of work stoppages may lead to project cost overruns.

The second impact of accidents is on the brand’s image. Media coverage of occupational accidents negatively affects the company’s reputation. Therefore, organizations that focus on brand image meticulously analyze the hazards that may attract media coverage.

Ultimately, risks can threaten society and the environment. The family of the victim may suffer financial/emotional damage, and co-workers may lose their motivation to work after an accident. Moreover, trivializing work accidents can make workers unresponsive to an unsafe working environment. The environmental impact of risks arises particularly in cases of fire, explosion, or chemical spillage. The risk impact may extend beyond the construction site, damaging neighbors or the environment.

Given these considerations, it is evident that limiting the severity parameter in risk assessment studies to mainly physical injury and using it as a singular metric detrimentally impacts the overall efficacy of the evaluation. Diversification of the severity parameter, based on all influenced factors, would significantly improve the coverage of the risk assessment.

2.2. Multi-Dimensional Safety Risk Assessment Approaches

Recently, academics have acknowledged the importance of conducting safety risk assessment studies from a multi-dimensional perspective and have begun formulating novel multi-criteria risk assessment methodologies [19,44].

AHP and the Analytic Network Process (ANP) are widely used MCDMs in safety risk assessment [21,36]. AHP addresses decision-making problems using a hierarchical framework, whereas ANP analyzes the interconnections in alternatives and criteria [55,56]. Liu and Tsai [57] introduced a fuzzy ANP methodology to identify significant hazard categories and their causes via network diagrams derived from two-stage Quality Function Deployment tables. Similarly, Aminbakhsh et al. [46] developed the AHP method as a tool to mitigate inconsistencies of risk severities in safety risk assessment. They also improved the effectiveness of risk assessment by decomposing the problem into more manageable sub-problems. Koulinas et al. [58] extended the AHP approach by incorporating the Proportional Risk Assessment Technique (PRAT) and enhanced it using historical data.

Another MCDM commonly utilized in the literature is the TOPSIS. Khan et al. [38] successfully ranked hazards in the construction industry using the Fuzzy TOPSIS method.

The literature also provides studies that employ AHP/ANP combined with other MCDMs. Rostamzadeh et al. [21] used the Decision-Making Trial and Evaluation Laboratory (DEMATEL) method combined with the ANP to identify and prioritize cause–effect linkages among factors causing construction falls. Mohandes et al. [19] introduced a novel methodology for risk assessment by combining the AHP with Fuzzy TOPSIS. Mohandes and Zhang [16] used Fuzzy TOPSIS and Fuzzy ANP algorithms to analyze construction hazards holistically.

Studies mainly focus on determining the priorities of risk types against each other. A few studies aim to optimize the cost–benefit relationship of measures. However, cost–benefit performance analysis plays a key role in the preparation of the project budget. Papazoglou et al. [23] argued that precautionary costs need to be integrated into optimization and developed a model utilizing a multi-objective evolutionary algorithm. One of the most significant theories in this regard is the Cost of Safety (COS) theory, which Chalosn [59] first proposed. The COS theory posits a theoretical equilibrium where the total costs of measures equal the total costs of injuries, indicating the optimal level of investment (Figure 1) [46].

MCDM is categorized into two primary types: multi-attribute decision making (MADM) and multi-objective decision making (MODM), based on the number of alternatives considered [60]. In MODM, the decision problem involves multiple conflicting objectives within a continuous decision space that must be concurrently optimized under a set of constraints [61], whereas MADM methods are utilized for discrete decision problems aimed at selecting an alternative from a finite array of options [62]. MADM is categorized into superiority methods (ELECTRE, PROMETHEE), distance-based methods (TOPSIS), value or utility function (AHP), and pair-wise comparison methods (AHP/ANP). Based on computation time and solution quality, MODM is classified into mathematical programming models (linear programming, goal programming) and heuristic algorithms (simulated annealing, GA) [60]. Therefore, in the context of construction safety risk assessment, the GA approach can be separated from MADMs (such as AHP, ANP, TOPSIS, ELECTRE, PROMETHEE, etc.) by specific characteristics: whereas these methods are utilized for discrete MADM problems aimed at selecting an alternative from a restricted selection of options, the GA approach addresses MODM problems that refer to multiple conflicting objectives within a continuous decision space requiring simultaneous optimization under a defined set of constraints.

MADM methods consider expert inference to rank the performance of various countermeasure alternatives. Such inferences can be situation/condition-specific and therefore difficult to generalize to different scenarios [63]. The MADM methods, such as AHP and TOPSIS, fail to effectively evaluate criteria and alternatives concurrently [11]. Nature-inspired optimization techniques such as the GA can be more efficient in non-linear optimization problems such as cost–benefit trade-offs [64]. For this reason, some studies have used artificial intelligence (AI)-based optimization algorithms to evaluate multi-criteria risks [11,23,65]. In contrast to conventional optimization methods, GA functions concurrently across several search areas rather than focusing on a singular point [65]. The integration of extra measures in risk management increases spending; however, the utilization of GAs enhances risk optimization, eliminating this issue [22].

2.3. Research Gap

Recent studies on Occupational Health and Safety (OHS) issues in the construction industry indicate several areas for improvement:

• Current research indicates that OHS risks impact not only the workers’ physical integrity but also the environment, society, reputation of the company, duration of projects, and costs (Table 2). However, a complete risk assessment approach that addresses all these issues is lacking. In this context, current risk assessment models require enhancements to incorporate these factors.
• The literature highlights the limits of approaches such as TOPSIS and AHP, which have been employed in recent years to assess OHS hazards. AI-based optimization techniques can significantly enhance the identification of optimal options by presenting broader solution spaces. Consequently, integrating AI-driven methodologies emerges as a significant domain for consideration in risk assessment procedures.
• The current literature reveals that comprehensive risk assessment methods are inadequate to reduce risk scores and precautionary costs for each work item in building projects. Such models should provide more comprehensive and cohesive strategies for reducing risks and costs. Developing models that optimize risks and costs for each work item is essential to improving the effectiveness of risk management practices.

2.4. Aims and Objectives

This study aims to assess the building construction risks by developing a risk assessment approach with multidimensional severity parameters (project cost, project duration, environment, society, and the company’s reputation).

Taking all identified precautions minimizes the overall risk score while increasing the prevention cost [23]. In the study, a multi-objective optimization model was developed that considers various risk parameters and maximizes the cost–benefit balance of measures. The developed model has been applied to a multistory building construction and the results were evaluated. This approach enables the development of OHS policies by ensuring the implementation of a cost-effective risk mitigation strategy that fulfills the specified risk criteria. Users will have the ability to determine cost-effective measures based on the unique risk criteria of their project.

3. Model Development and Research Methodology

The proposed model consists of five main steps. The steps are given in Figure 2.

Step 1: Identification of risks

The risks and precautions were identified by a literature review and risk assessments from completed building construction projects.

Step 2: Constructing the risk assessment form

The discovered risks were classified by work items, and a risk assessment form was developed to assess probability (P) and severity (S) scores. The determined severity parameters include the impact on workers (which, for the sake of the terminological integrity of the proposed approach, can also be referred as Impact on Worker Health and Safety (IoWHS), the impact on the environment (IoE), the impact on society (IoS), the impact on project cost (IoPC), the impact on project duration (IoPD), and the impact on the reputation of the company (IoRC). The severity parameters and description are given in

Table 3. Severity parameters and descriptions.

Parameter	Description	Parameter	Description
P	Degree to which the risk is likely to materialize	IoPD	Impact of occupational accidents on project duration
IoWHS	Impact of risk on workers health	IoPC	The impact of cost increases such as lawsuits that may be filed against construction companies affected by accidents, compensation paid to the family of the injured worker
IoE	Impact of fire, explosion, or vehicle accidents on the environment or bystanders in situations
IoS	Moral impact of accidents on society in companies, victims’ families and colleagues, or society in general	IoRC	The impact of media coverage of accidents or accusations made by victims or their families

.

Step 3: Peer-Review of Risks

The risk assessment forms were distributed to safety professionals experienced in the risk assessment of building constructions.

Step 4: Optimization Module and User Interface

Following the determination of risk scores by experts and the investigation of the costs of the measures, cost–benefit optimization was performed using a population-based optimization algorithm.

Step 5: Case Study

The proposed model was applied to a 21-story building construction. The results were compared with the previously implemented traditional risk assessment results.

3.1. Identification of Risks

The initial phase of the study is to identify the risks and prevention alternatives in building construction. Construction projects are unique, and risks may vary according to the characteristics of the project and the environment. For example, the structural system of a building project may differ between reinforced concrete and steel construction, which may lead to completely different hazards and risks encountered during construction. Also, the existence of a high-voltage line at the construction site poses additional electric shock risks, whereas they are inherently mitigated at sites devoid of such infrastructure.

In this study, the risks and prevention alternatives associated with multistory reinforced concrete building construction projects were identified through a review of the literature and risk assessment cases. Research articles and conference proceedings addressing risks and measures in building constructions were analyzed using databases including Google Scholar, ScienceDirect, Web of Science, Springer, Taylor and Francis, and Wiley. The literature review utilized the terms “construction industry”, “occupational safety and health risks”, “building construction”, “safety cost”, “safety measure”, and “risk assessment.” Following the review, 27 research articles and conference papers were selected.

Risk assessment cases were analyzed to identify the main risks related to reinforced concrete building construction. Risk assessments of ten multistory reinforced concrete building construction projects for small and medium-sized enterprises were analyzed, and the main risks and safety measures for each work item were identified (

Table 4. Hazards/risks and potential precautions.

Work Items	Hazards/Risks	Precautions
Concrete	Concrete mixer hits workers, concrete mixer hits surrounding structures/vehicles, electric shock, fall from height	Having a marker person on the field, insulated gloves, warning signs, and safety rope systems
Covering	Material falls, improper use of hand tools, electric shock, the air well remains open	Railing systems, building cover systems, protective goggles, warning signs, insulated gloves, closing of air well
Electrical wiring	Electric shock, fire risk	Insulated gloves, warning signs
Excavation	Fall from height, construction equipment hitting people and objects, entering dangerous areas, electric shock, the vehicle overturned at the excavation site, workers were trapped under the cave-in, inhalation of dust, occupational accidents due to machine malfunctions, improper use of hand tools, hitting the power transmission lines	Railing systems, protective goggles, warning signs, insulated gloves, closing of air well, security area curtains, having a marker person on the field, safety lane, dust mask
Formwork	Splattered materials hitting workers, material fall, improper use of hand tools, fall from height	Railing systems, safety rope systems, protective goggles, security area curtains, safety lane
Masonry andchimney	Fall from height, material fall, the air well remaining open, improper use of hand tools, electric shock, inhalation of dust, object getting into the eye	Railing systems, safety rope systems, protective goggles, warning signs, insulated gloves, closing of shaft gaps, building cover systems, dust mask
Mechanical installation	Falling through elevator shafts, improper use of hand tools, dust getting into the eye, inhalation of dust, material fall, electric shock, the air well remains open	Railing systems, safety rope systems, protective goggles, warning signs, insulated gloves, building cover systems, closing of air well, dust mask
Insulation	Fall from height, material fall, improper use of hand tools, electric shock, the air well remains open	Railing systems, safety rope systems, protective goggles, warning signs, insulated gloves, building cover systems, closing of air well, dust mask
Joinery	Fall from height, electric shock, improper use of hand tools, material fall, the air well remains open	Railing systems, safety rope systems, protective goggles, warning signs, insulated gloves, closing of air well, building cover systems
Plaster and paint	Improper use of hand tools falls from height, material falls, inhalation of dust, objects getting into the eye, poisoning, the air well remaining open	Railing systems, safety rope systems, protective goggles, warning signs, building cover systems, closing of air well, dust mask
Rebar works	Improper use of hand tools, electric shock, fall from height	Railing systems, safety rope systems, protective goggles, warning signs, insulated gloves
Roof	Fall from height, material fall, improper use of hand tools, electric shock, the air well remains open	Railing systems, safety rope systems, building cover systems, protective goggles, insulated gloves, warning signs, closing of air well

). The findings were combined with the literature and the main risks and alternative precautions were determined. Ultimately, the unit costs of the measures have been determined by analyzing the average prices offered by a minimum number of three national suppliers. The unit costs of some of the precautions to be implemented in the project are as follows: safety rope: USD 56.26, safety goggles: USD 0.94, and safety tape: USD 0.01.

3.2. Constructing Risk Assessment Form

Following the identification of safety risks and alternative measures, a risk assessment form was designed for safety professionals to assess. While various risk assessment methods (like matrix-type, Fine–Kinney, FTA, and FMEA) can be applied in construction-related risk assessment forms. The matrix-type assessment approach has been preferred due to its ease of use [32,66]. However, some modifications were applied to the matrix-type assessment approach. In the traditional method, the risk score is calculated by multiplying P and S; instead, the suggested model employs a multidimensional approach that multiplies P by the IoWHS, IoE, IoS, IoPC, IoPD, and IoRC. In the calculations, each parameter was scaled with a five-point Likert scale in parallel with the classical method.

The risks were listed on a work item basis in the assessment form, with corresponding precautionary alternatives specified for each work item.

Table 5. Risk assessment form (sample—roofing activities).

Roofing Activities	P	IoWHS	IoE	IoS	IoPC	IoPD	IoRC
Fall from height (No precautions)	…	…	…	…	…	…	…
Fall from height (Railing systems)	…	…	…	…	…	…	…
Fall from height (Safety rope systems)	…	…	…	…	…	…	…
Material fall (No precautions)	…	…	…	…	…	…	…
Material fall (Railing systems)	…	…	…	…	…	…	…
Material fall (Building cover systems)	…	…	…	…	…	…	…
Improper use of hand tools (No precautions)	…	…	…	…	…	…	…
Improper use of hand tools (Protective goggles)	…	…	…	…	…	…	…
Electric shock (No precautions)	…	…	…	…	…	…	…
Electric shock (Insulated gloves)	…	…	…	…	…	…	…
Electric shock (Warning signs)	…	…	…	…	…	…	…
Falling of materials through open shaft voids (No precautions)	…	…	…	…	…	…	…
Falling of materials through open shaft voids (Closing of air well)	…	…	…	…	…	…	…

presents an illustration of a section of the risk assessment form on roofing activities.

3.3. Peer-Review of Risks

Risks were peer-reviewed by professionals specialized in occupational safety in construction. While many suggestions are available on sample adequacy in expert opinion research, related literature recommends the optimal number of specialists to be in the range of 10 to 20 [67,68,69,70]. Moreover, analogous research considered criteria such as experience, professional qualifications, and educational background in the identification of experts [68]. Therefore, the risk assessment form developed in this study was completed by 10 specialists from the risk assessment team who are experts in reinforced concrete building structures, utilizing both in-person and online techniques. All the experts have qualifications as occupational safety experts and more than five years of construction safety risk assessment experience. The main demographics of experts are presented in

Table 6. Demographics of the experts.

Position	Professional Experience (Years)	Experience in Risk Assessment (Years)	Education
Academician	50	20	Doctoral Degree
	15	10	Doctoral Degree
	17	10	Doctoral Degree
Occupational Safety Expert	17	7	Master’s Degree
	5	5	Master’s Degree
	7	5	Master’s Degree
	8	5	Master’s Degree
	10	5	Master’s Degree
	7	5	Bachelor’s Degree
Civil Engineer	20	20	Master’s Degree

.

3.4. Optimization Module and User Interface

Optimization is the process of utilizing a situation or resource as efficiently as possible [71]. Heuristic and classical optimization methods are the two primary subcategories of optimization algorithms. Classical optimization algorithms are useful for finding the optimum solution, or the maximum and minimum values of continuous or differential functions. Algorithms that converge to near-optimal solutions in the solution space are known as heuristic optimization algorithms. Heuristic optimization algorithms are usually inspired by the behavior of animals, and physical and natural phenomena to address complex problems [72,73,74]. The GA, Firefly Algorithm (FA), Artificial Bee Colony Algorithm (ABC), Flower Pollination Algorithm (FPA), Ant Colony Algorithm, and PSO are a few of the most used heuristic algorithms [75]. In this study, due to the binary nature of the precaution selection process (1: Select/0: Not Select), a solution can be created as a binary set of precautions called a string, whose length depends on the number of work items and their associated risks. Thus, the GA, which is one of the most widely used binary optimization methods, was employed in this research and its implementation details and optimization results are reported.

The Binary Firefly Algorithm (B-FA), Binary Artificial Bee Colony Algorithm (B-ABC), and Binary Flower Pollination Algorithm (B-FPA) were additionally integrated into the optimization module to provide flexibility to the user for experimenting with different algorithms. Furthermore, the system was developed in such a way that the module can be further extended to work with other optimization algorithms that operate on binary solutions.

All the aforementioned binary optimization algorithms were implemented and tested within the optimization module. Their performance was comparable to that of the GA, as they all converged to the same optimal solution. However, since this study focuses on developing a multi-dimensional construction-safety risk optimization model, the implementation details of B-FA, B-ABC, and B-FPA are not included. The GA implementation is presented in this paper not due to its superiority over the tested algorithms but because it is a well-established and widely recognized algorithm among researchers.

3.4.1. Implementation of the Genetic Algorithm

The GA is a heuristic optimization method that is inspired by the principles of evolution. The algorithm operates through the iterative application of selection, crossover (recombination), and mutation on a population of candidate solutions. Constructing a candidate solution in the form of a chromosome is a crucial requirement of the GA. Each solution, often encoded as a string of genes (typically in binary), represents a potential answer to the problem. The algorithm evaluates the fitness of each candidate based on a predefined objective function, which is called the “fitness function”. The first step of the GA is to create a collection of solutions randomly that is called the “initial population”. Then, the GA creates a new population by applying crossover and mutation operations on the promising existing individuals, which are called parents. This process is repeated iteratively through the generations hoping that the individuals in each successive generation will be improved versions of their ancestors. The algorithm terminates if the optimum solution is found or the predetermined number of generations is reached. In the latter case, the best solution ever found can be the optimum or near-optimum solution to the problem. The method is especially effective in complex, multi-dimensional search spaces where traditional algorithms might struggle to find optimal solutions efficiently.

The initial stage in the optimization process is to determine the work items and severity parameters to be optimized. By selecting these work items and impact types via the user interface, binary control strings called “tickets” that allow the optimization module to work flexibly are created. The “tickets” are used by the function that evaluates how good the created solution is by taking into account only the selected impact factors and work items. The user interface issues another ticket that is called “Work Items Ticket” to inform the optimization module how to build up chromosomes and to inform the fitness function of which precautions will affect the quality of the solution (Figure 3). Furthermore, by selecting the optimization algorithm and its related parameters, a final ticket, the “Algorithm and Settings Ticket”, is issued by the user interface module. All those so-called tickets govern the optimization module and provide encapsulation and extensibility to the modules.

The next step is to create an initial population, which is the collection of candidate solutions. A candidate solution is represented as a binary chromosome whose genes are the decision parameters that control the actions to be taken for related precautions (Figure 3). Each risk (p_i) is represented with a variable length binary code and located in a relative position in the chromosome segment of the related work item. The relative positions are determined from the information that is provided by “Work Items Ticket”. The Chromosome Aggregator attaches the work items chromosome segments whose ticket values are 1 in a predefined order. The output of the Chromosome Aggregator is a binary string that contains the precaution codes of selected work items.

A chromosome, which is also called an individual, is created randomly by filling 1/0 of its genes. The GA is a population-based algorithm. Therefore, a collection of individuals is created randomly and is called “initial population”. The GA aims to improve the quality of the population by creating new individuals from the existing ones. The quality of individuals is determined with the fitness function $f\left(\mathsf{\omega },T\right)$ that is given in Equation (6). The derivation of fitness function is explained as follows.

Participants were requested to assess the P, IoWHS, IoE, IoS, IoPC, IoPD, and IoRC, contingent upon the preventative measures for identified risks related to work items. The “mode” of responses of the experts to each item was then assigned as the consensus score of that risk. In this way, risk scores were obtained by multiplying the severities and probability parameters. These scores were then summed as in Equation (1) to obtain the total risk value for all work items. Finally, the risk value is normalized using Equation (2).

$$TotR(\omega ,T)=\sum _{j \in \omega }\sum _{i=1}^{{NoR}_j}{P}_{i,j}\ast \left({IoWHS}_{i,j}+{T_e\ast IoE}_{i,j}+{T_s\ast IoS}_{i,j}+{T_{pc}\ast IoPC}_{i,j}+{T_{pd}\ast IoPD}_{i,j}+T_{rc}\ast {IoRC}_{i,j}\right)$$

(1)

$$R_{Total}^{*}(\mathsf{\omega },T)=\left({\displaystyle \frac{TotR\left(\omega ,T\right)-R_{min}}{R_{max}-R_{min}}}\right)\ast 25\hspace{1em}\hspace{1em}\hspace{1em}(0\le R_{Total}^{*}(\mathsf{\omega },T)\le 25)$$

(2)

where ω and T are the set of works and impact types to be analyzed, respectively. ${NoR}_j$ is the total number of risks in the work type j, $P_{i,j}$ is the probability of a risk, ${IoWHS}_{i,j}$ is the severity of a risk, ${IoE}_{i,j}$ is the impact of a risk on the environment, ${IoS}_{i,j}$ is the impact of a risk on the society, ${IoPC}_{i,j}$ is the impact of a risk on the project cost, ${IoPD}_{i,j}$ is the impact of a risk on the project duration, and ${IoRC}_{i,j}$ is the impact of a risk on the company’s reputation. $T_e, { T}_s,{ T}_{pc},T_{pd}, \mathrm{a}\mathrm{n}\mathrm{d} T_{rc}$ are the impact control variables whose values (0 or 1) are assigned by the tickets issued in the user interface. TotR($\mathsf{\omega },T$) is the sum of the risks determined based on the work items to be analyzed, $R_{Total}^{*}(\mathsf{\omega },T)$ is the normalized total risk value, $R_{min}$ is the lowest value within the risk values, and $R_{max}$ is the highest value within the risk values.

The costs of collective and personal protective precautions against risks were calculated as discussed in Section 3.1 and then normalized using Equation (3) to ensure that big differences between the costs of precautions do not lead to biased results. Using these cost values, the total cost value of the precautions was calculated using Equation (4).

$${ m}_i^{*}=(R_{max}-1)\ast (1-e^{-{\displaystyle \frac{1}{2{(\frac{{ m}_i-m_{min}}{\tilde{m}-m_{min}})}^2}}})+1 \hspace{1em}\hspace{1em}\hspace{1em}(1\le m_i^{*}\le 25)$$

(3)

$$TotC(\mathsf{\omega })={\sum }_{j \in \mathsf{\omega }}{\sum }_{i=1}^{{NoP}_j}{m}_i^{*}\ast I_i$$

(4)

where $m_i^{*}$ is the normalized cost value, ${NoP}_j$ is the total number of precautions in work item j in the work set ω, $R_{max}$ is the maximum risk score (5 × 5 = 25), $m_i$ is the cost of the precaution ($), $m_{min}$ is the lowest value among the cost values, $\tilde{m}$ is the median of the cost values, $TotC(\mathsf{\omega })$ is the sum of the cost values of the precautions, and $I_i$ is the value (1 or 0) of the precautions.

Equation (5) is used to normalize the total cost.

$$C_{Total}^{*}(\mathsf{\omega })=\left({\displaystyle \frac{TotC\left(\mathsf{\omega }\right)-C_{min}}{C_{max}-C_{min}}}\right)\ast 25 \hspace{1em}\hspace{1em}\hspace{1em}(0\le C_{Total}^{*}(\mathsf{\omega })\le 25)$$

(5)

where $C_{Total}^{*}(\mathsf{\omega })$ is the normalized cost, $C_{min}$ is the lowest value within the costs, and $C_{max}$ is the highest value within the costs. Then, fitness function f(), is calculated using Equation (6).

$$f\left(\mathsf{\omega },T\right)={{{(R}_{Total}^{\ast }(\mathsf{\omega },T)}^2+{C_{Total}^{\ast }\left(\mathsf{\omega }\right)}^2)}^{1/2}$$

(6)

The GA replaces the existing population with a newly created one in each consecutive iteration, which is called a “generation”. A new population is created by selecting some fit individuals that are called “parents” and obtaining new individuals that are called “offspring”. There are many strategies in the literature for the selection of parents. The Roulette Wheel, Tournament, and Ranking Selection methods are used widely in the literature. In this study, the Roulette Wheel is used as a selection method. Once a parent is selected, it can either be directly transferred to the next generation or crossover with another parent. This is determined with “crossover probability”. In this study, Single-Point Crossover, which is a very popular method for the recombination process of binary chromosomes, is employed. The offspring creation process is completed after applying mutation operation. Some of the genes of an offspring are selected with a mutation probability, and they are randomly modified. Flip-Bit mutation that replaces a gene with its complement is very effective on binary chromosomes and, therefore, is employed in this work. Furthermore, some of the fittest individuals are directly transferred into the next generation in order to preserve the best individual created. This strategy is called elitism, and the number of elites is set to 1 empirically. The best chromosome that reduces the precaution cost and risk score is identified after running the GA with a predefined maximum number of generations. The population size, maximum number of generations, crossover, and mutation probability values have to be set before running the algorithm.

Grid search is employed for parameter selection in each of the tested algorithms. Experimental results reveal that a moderate population size of 50 consistently yields outstanding performance across all tested parameter combinations. While all algorithms successfully identify the optimal solution within a few seconds, variations in convergence time are observed.

Figure 4 presents heatmaps illustrating the convergence times of the algorithms to the optimal solution across all tested parameter configurations. Since all tested settings provide the optimal solution, early convergence time has been considered the primary criterion for parameter selection. The selected settings are marked with ‘*’ on the heatmaps. Based on these heatmaps, we recommend using the algorithm parameters listed in

Table 7. Recommended parameter settings for the algorithms.

Parameter	Optimization Algorithms
	GA	B-ABC	B-FA	B-FPA
Population size	50	50	50	50
Number of generations	150	150	150	150
Elit count	1	--	--	--
Crossover probability	1	--	--	--
Mutation probability	0.2	--	--	--
Scout probability	--	0.4	--	--
Switch probability	--	--	--	0.5
Phi	--	--	--	0.9
Initial Gamma	--	--	1	--
Beta	--	--	0.7	--

.

To ensure a fair comparison, the same population size and iteration numbers are used for all tested algorithms. However, it should be noted that every project poses different challenges to the algorithms. Therefore, it is recommended to experiment with the parameters to identify the best-performing settings according to the project’s needs.

3.4.2. User Interface

An interface has been constructed to provide users with greater flexibility during the assessment process (Figure 5). The interface allows users to select the optimization algorithm and perform analysis on the desired combination of work items and severity parameters. The user interface and data analysis were executed utilizing the Python 3.10 environment. The main libraries used are as follows: Mkl_Random 1.2.4, Matplotlib.3.4.8, Deap 1.4.1, PySimpleGUI 5.0.6, and Mpmath. 1.3.0

3.5. Case Study

Occupational accidents in the construction industry mostly occur on small and medium-sized projects [28]. Thus, a medium-sized building construction project was preferred to validate the proposed model. This project, located in Türkiye, is a 21-story reinforced concrete structure with a 10,350 m² total building area. The project consists of sixty-three apartments, two playgrounds, and a shelter. The project includes work items such as excavation, formwork, concrete, rebar works, mechanical and electrical installations, masonry works, joinery, covering, plaster and paint, roofing, and insulation. In addition, construction equipment such as concrete mixers and excavators were used on the site.

Through the risk analysis conducted by the project’s risk assessment team using the traditional L-type matrix method (RS = P * S), project-specific risks and precautions to be taken were analyzed. These risks and precautions coincide with the ones in Table 3. Thus, through the proposed risk assessment method, risk scores and costs were calculated according to cases where no precautions and all precautions were taken. In these calculations, the significance of each severity parameter is presumed to be equivalent. While making the cost calculations, the architectural details of the building and the number of workers working on the relevant work items were considered. Then, the unit costs of the precautions calculated according to the method described in Section 3.1 were multiplied by these values to obtain the total cost of each precaution for this construction site. Finally, the results of the proposed approach are compared in terms of prevention costs and risk scores between the cases where no precautions are taken and all precautions are taken; the results of the comparison are presented below.

4. Results and Discussion

The first phase of validating the proposed optimization model is selecting the work items and severity parameters. All work items and severity parameters were initially selected to validate the model, followed by testing various combinations of severity parameters to illustrate the model’s flexibility.

Figure 6 illustrates the total cost, total risk, and best fitness values for the first 20 generations for ease of readability. The figure demonstrates that the GA successfully reaches an optimal solution, identifying the best solutions based on total cost and risk factors by the 12th generation. While risk was minimized in this process, fluctuations in the cost of measures were observed. These fluctuations indicate that the cost initially stayed high and then became more consistent as the risk decreased. The main purpose of risk assessment is to eliminate and/or control hazards with critical and high potential risks to the lowest reasonable level of risks to protect employees [28,76]. The model results support this theory.

It should be kept in mind that optimum investments to prevent risks will reduce or limit potential accident costs [77]. These investments can significantly reduce the overall costs associated with occupational health and safety in projects [78]. Proper investments enhance safety levels while simultaneously reducing long-term operating costs. In summary, preventing accidents and minimizing risks in the long term with these precautions both ensure the safety of employees and contribute to the economic sustainability of the business. The results show that the risk score and the cost of precautions have decreased significantly and are approaching an optimal equilibrium (Figure 6). This equilibrium was attained by the efficient implementation of precautions and cost optimization. Thus, it illustrated the relationship between security and economic efficiency.

Table 8. Total risk scores and precaution costs.

Precautionary Status	Risk Scores							Total Cost ($)
	IoWHS	IoE	IoS	IoPC	IoPD	IoRC	Total
No precautions	986	748	711	726	700	780	4651	0
Proposed model	467	257	287	266	270	275	1822	29,178.89
All precautions	325	162	188	178	173	181	1207	51,734.08

presents the results of the case study with all work items and severity parameters. Traditional Risk Score (TRS) is derived by multiplying the P and IoWHS, akin to conventional matrix-based risk evaluations. The risk parameter scores (IoE, IoPD, etc.) were derived by multiplying the established severity parameters by P. Consequently, the risk scores for IoWHS, IoE, IoPC, IoPD, IoS, and IoRC rose by 21.24%, 16.21%, 16%, 18.40%, 18.92%, and 15.69%, respectively, in comparison to the situation encompassing all precautions (Table 8). The precaution cost “all precautions” represents the cost of taking all the precautions identified in the risk assessment. As a result of the optimization, it was concluded that eight of the identified precautions should be taken to reach the optimum level of risk score and cost. These precautions are “having a marker person on the field”, “warning signs”, “safety ropes”, “building cover systems”, “protective goggles”, “dust masks”, “safety area curtains”, and “safety tape”. Forteza et al., [77] emphasize that over-precautions negatively impact companies financially. Therefore, it is crucial to assess the cost–benefit of precautions to avoid economic damage to firms [46]. To assess the efficacy of the measures relative to their costs, the expenditure was divided by the risk score it mitigated, and the cost for reducing one unit of risk (unit cost of benefit—UCB) was calculated. While the UCB in the proposed model is USD 10.31, it is USD 15.02 in the risk assessment prepared for the project.

Comparing the developed model to the traditional risk assessment approach, i.e., taking all precautions, the risk score increases by 17.86% (1822-1207 = 615 points), and the cost of precautions decreases by 43.60% (51,734.08-29,178.89 = $22,555.19). It is well known that construction companies tend to neglect taking necessary precautions and only begin spending when accident rates reach exceptionally high levels [77]. Hence, it is crucial to develop risk assessment methodologies that demonstrate the feasibility of preventing accidents at low costs [79]. Thus, although the unfavorable 17.86% rise in risk score, the result could potentially motivate enterprises, particularly in developing countries, who perceive occupational safety precautions as unnecessary expenditures [28]. Paying 57.40% of the preventive cost can lead to an 82.14% reduction in the risk score. In addition, sharing impact scores with stakeholders will contribute to raising awareness and breaking decision-makers’ resistance to take precautions, as it will demonstrate the consequences that may be encountered. This enables workers to work in a more salubrious atmosphere.

The findings of the suggested model indicate that suitable measures are implemented for all risks, except for the risk of objects descending from height due to the open-air shaft and the risk of falling from height during excavation. The TRSs of the accepted risks are 20 and 25, which are categorized as “high risk” and “intolerable risk”, respectively [80,81]. The model does not yield perfect solutions, and the results require verification by professionals. Alternatively, various constraints can be defined in the model to avoid ignoring unacceptable risks.

Even though the severity parameters of IoWHS, IoE, IoS, IoPC, IoPD, and IoRC have been defined in the study, decision-makers may not always prefer to perform assessments by considering all these parameters. As mentioned before, in the proposed model, the dimensions of severity are designed to be user controllable. Thus, professionals can evaluate by considering the parameters of their choice.

Table 9. Total risk scores categorized by various severity combinations.

	IoWHS	IoWHS, IoE	IoWHS, IoS	IoWHS, IoPC	IoWHS, IoPD	IoWHS, IoRC	IoWHS, IoPC, IoPD	IoWHS, IoE, IoS, IoRC	All Parameters
Total Risk Score	462	714	746	733	737	742	1003	1286	1822
Total Cost ($)	29,179.47	29,179.47	29,179.47	29,179.89	29,179.89	29,179.89	29,179.89	29,179.89	29,179.89

presents the optimization results using different severity combinations.

The results show that the optimum cost is achieved in all combinations for the current project (Table 9). The linear relationship of the severity parameters resulted in the convergence of the optimal solution. However, it is not possible to assume that the optimum solution will always occur at a single point as the ideal approach may vary based on the selection of severity parameters in projects that involve several risk categories.

5. Implications

This study proposes a novel approach for safety risk assessment in building construction projects. The model offers a more systematic and comprehensive approach to evaluating risks in construction projects. The novel multidimensional assessment model guarantees that safety risks on the construction site are assessed using various severity parameters and are optimized with prevention costs. It enables risk prioritizing based on overall impact rather than solely on the probability of injury. This approach offers various implications in practice and theory.

The classical L-type assessment approach relies on a singular severity parameter, which solely quantifies the severity of the risk based on the physical injury experienced by the worker. A work accident at the construction site jeopardizes not only the health and safety of the workers but also adversely impacts the project’s success, the organization, and the environment. The novel approach combines physical risk with project, social, and environmental risks. Consequently, hazards can be addressed from multiple perspectives and evaluated thoroughly. Thus, assessors can examine risks from a wider perspective. Furthermore, the developed model allows the selection of both severity parameters and activities on a work-item basis, enabling optimization in preferred combinations. Safety professionals may modify weightings according to project priorities, legal requirements, and stakeholder expectations. It is well known that all construction projects have unique risks, with the degree of severity of each risk potentially differing based on the specific project. Most construction projects, including dams, hydroelectric plants, and coastal structures, are governed by environmental and social regulations. The approach contributes to providing enhanced compliance with legal requirements by evaluating risks from environmental and social perspectives. This further helps an organization in achieving sustainable construction goals. Moreover, the flexible nature of the model allows for easy adaptation to engineering structures. Given the characteristics of engineering structures, different severity parameters may be prioritized. The flexible structure of the model allows it to be adapted to different risk groups and allows the weights on severity parameters to be easily adjusted.

Construction projects are planned with a budget and time constraint and implemented following these goals. Safety risks may result in project interruptions and various types of compensation. Assessing safety risks from the perspective of project duration and budget can contribute to meeting time and budget targets.

Occupational accidents at construction sites can negatively affect the reputation of the company and damage the trust of stakeholders. Addressing the negative impact of risks on the company’s image will increase confidence in stakeholders, including subcontractors and customers.

Another weakness of classical assessment methods is the lack of cost–benefit analysis of the measures. The impact of the cost of the measures on risk minimization has not been considered. Numerous construction companies encounter financial constraints, resulting in challenging decisions regarding safety expenditures, especially in developing countries. This requires that contractors maximize the efficacy of safety investments. The proposed optimization methodology offers significant potential for contractors in developing countries that are economically forced to do business with high efficiency. Understanding the relationship between safety investments and benefits and establishing equilibrium between these variables will directly influence project success. This methodology offers cost–benefit optimization that ensures the optimal level of risk is achieved at the lowest possible cost. Optimizing risk scores and safety-precaution costs, which facilitates the effective allocation of resources to address the most significant hazards, ensures both effectiveness and cost efficiency. The model searches for the best tradeoff between total risk scores and safety-precaution costs. This means that safety investments yield the maximum benefit per amount invested. The optimization model reduces unnecessary expenditure while meeting safety requirements. Thus, the optimization of safety expenditures offers a competitive advantage.

The model also possesses the capability for integration into Building Information Modeling (BIM), project management platforms, or safety management systems. The model can be integrated into any IT system via an API-based interface, offering efficient application in projects.

6. Limitations and Future Directions

This study also has some limitations. The proposed model is applicable for various project types; however, it has been tested solely within the primary risk groups of building projects. However, construction projects are not limited to building projects but also include many engineering structures such as tunnels, bridges, and dams. Engineering structures are often more complex than building projects and involve unique risks. For instance, the main source of uncertainty in a dam or tunnel project is geotechnical conditions. These projects pose specific risks, such as large cave-ins and poisoning from underground gases. Thus, the model requires validation for risk categories of several civil engineering projects, such as tunnels, dams, road construction, etc., and its applicability must be verified. Furthermore, integrating various severity parameters into the model may diminish the significance of physical injury which should be the top priority. To address this limitation, parameters can be weighted using various approaches, such as AHP, MOORA, etc., according to the organization’s priorities. During the optimization process, the failure to implement protections against a few unacceptable risks is observed. This is thought to be related to the algorithm’s mathematical focus on finding the optimum solution. Therefore, it is very important that the appropriateness and applicability of the results produced by the model are checked by experts. To solve this problem, it is also suggested to use hybrid and other up-to-date algorithms or define specific risk score constraints. In addition, after the outputs of the model are evaluated by experts, expert feedback should be integrated into the model and then the outputs should be implemented. As a result, the developed model does not provide perfect results but allows experts to select appropriate measures with optimum efficiency.

The model is also likely to face some challenges in its implementation. Especially in large-scale projects, many precautionary alternatives for thousands of risks need to be evaluated. This increases the search space and computational cost of the algorithm considerably. The adoption of cloud-based computing techniques could accelerate the model. Moreover, large-scale projects may include different subcontractors, each adhering to their own safety protocols. The model’s centralization to provide access for all subcontractors may enhance the coherence of safety practices.

A further limitation of the approach is the lack of real-time risk assessment. The model can be updated by dynamically integrating environmental and project-specific conditions using real-time construction monitoring systems.

7. Conclusions

Risk assessment is an essential task for preventing occupational accidents in the construction industry. Hence, it is imperative to conduct an in-depth assessment, considering every possible effect of the risks. However, previous studies typically assess risks based on their potential impact on workers. On the other hand, the lack of understanding of stakeholders on occupational safety hinders the implementation of appropriate precautions at construction sites. In parallel, studies on workplace safety precautions have primarily concentrated on assessing the costs of precautions and their proportion to project expenses. Notwithstanding all efforts, there has been no reduction in overall occupational accident rates in the construction industry. This suggests that there is a need to develop innovative approaches to enhance stakeholder awareness.

This study develops a model that identifies and minimizes the risks and associated costs in building construction. The model focuses on determining precautions on a work-item basis to reduce both the risk score and the precaution cost. Contributing to the body of knowledge, this study addresses a multi-objective problem in which the cost of precautions is considered, and risk scores are calculated based on work items. In addition to the commonly used probability and severity parameters, risk scores were also calculated by considering the severities of impact on the environment, project cost, project duration, society, and company’s reputation. In this way, the effects of risks were evaluated with different severity parameters, and precautions were determined by considering risk scores and prevention costs. The GA was employed to address the multi-objective problem, and a typical multistory building project was used as a case study to assess the model’s success. Comparing the project’s results with classical risk assessment, it was determined that while there was a slight increase in the risk score, there was a significant reduction in the cost of the precautions. A major challenge encountered by stakeholders in the industry is the neglect of safety precautions owing to financial constraints. The results indicate that effective health and safety implementation strategies can be developed even with limited budgets. It also enables more comprehensive results to be obtained through the parameters established. It is thought that proactive precautions can be taken to prevent accidents with an optimum risk score and cost practically by using the proposed model. The model offers users the ability to define GA configurations, work items, and severity factors, hence granting flexibility for academics and practitioners to evaluate risks that can be customized to their needs. The model also assists sector stakeholders in ascertaining the expenses associated with precautions during the planning phase of construction projects. In addition, the results obtained can be shared with stakeholders during consulting or training services to raise awareness.