BIM-Based Safety Leading Indicators Measurement Tool for Construction Sites

Recognition of safety leading indicators is considered as an important part of safety management in construction projects to prevent hazardous accidents occurrence. In the available literature, the lack of a comprehensive proactive approach for assessing safety, based on active safety leading indicators is tangible. In this study, BIM and knowledge base are integrated to develop a safety management framework, then, this framework is implemented in a case study to assess its performance. Safety leading indicators are extracted from experts’ expertise, documents, and best practice, and then the relationships among safety leading indicators, activities, and their attributes are established; these pieces of knowledge are stored in a safety leading indicator knowledge base. To measure the impacts of these indicators on the project, a framework is developed to integrate the knowledge base and building information modeling (BIM). The developed add-on generates a heat-map to visualize the impacts of the safety leading indicators on a daily basis, and it also produces a textual result sheet. To verify the effectiveness and efficiency of the developed tool a survey from professionals is conducted. This tool provides a way to automatically assess the impacts of the safety leading indicators. Using its results, safety managers and project participants can make important decisions to reduce hazards. The results also indicate that accidents occurring in the construction site dropped by 38% after employing the presented framework.


Introduction
The construction industry has always been considered as one of the most dangerous industries as it can be faced with various hazardous factors [1,2]. Each year, more than 60,000 fatal incidents occur in construction sites worldwide [3]. Although many attempts have been made to improve construction safety particularly in the last years, the construction industry is still ranked among the highest risky industries; much more attention is hence required to alleviate this issue [2]. Past studies have shown that each accident occurring in a construction site originates from one or several drivers, which are recognizable prior to the main accident [4][5][6]. This highlights the need for adopting a reliable approach in construction safety management systems and focusing on the origins of hazardous incidents. It has been discussed in many research that there is a correlation between the final incidents and the situation or prior incidents which facilitate their happening [7,8].
The concept of safety leading indicators has been brought up as conditions, events, or measures that precede an incident and have a predictive value regarding an accident, incident, or unsafe conditions [9]. Occupational Safety and Health Administration (OSHA), experience modification rating (EMR), recordable injury rates, days away, and restricted work, or transfer (DART) are the metrics that have been used traditionally for measuring the safety performance [10]. These traditional metrics which are known as lagging indicators mostly which rely on historical information from occurred accidents and incurred losses [11]. Utilizing leading indicators for measuring safety performance provides a proactive approach in safety management systems versus traditional passive approaches [10]. While the trend of using leading indicators has shown great potential for improving safety performance, some barriers have hindered its adoption in safety management systems, such as a tendency to use statistical sensible data by safety managers, lack of efficient tools and knowledge, and safety regulations in the construction industry which prevent innovative solutions [12,13].
Checklists are the initial attempts for building databases of leading indicators that are often formed based on empirical data or best practices [14]. As each project is carried out through unique situations, checklists containing leading indicators of safety hazards that are extracted from documentations and previous studies together with the knowledge of experts form a holistic and dynamic view about the issue during the project lifecycle [15]. Although these databases have an important role in the decisionmaking process in safety management, novel tools need to be utilized to improve the performance of such processes. On the other hand, advances in Information Technology (IT) during the last decade and its collaboration with safety management systems have facilitated the experts' judgment. These technologies have provided risk assessment meetings with higher productivity for project participants [16].
Building information modeling (BIM) provides a digital representation of the actual characteristics of elements in a project, which is a proper platform for sharing information among different stakeholders [17]. Moreover, BIM enhances the collaboration and integration level through projects. BIM has shown considerable advantages in projects' lifecycle from early stages, such as project definition and design, during construction and operation phases, and until the demolition process at the end [18].
In previous studies, the relationship between safety leading indicators and the safety performance of the projects has been evaluated [19]. Furthermore, knowledge bases for safety leading indicators have been established and allocated to projects in a holistic view, not to the activities [20]. Moreover, the relationship between upcoming activities and the recorded injury reports and other safety factors have been investigated; this asserts the relationship between passive safety leading indicators and the activities of the projects [21]. Even though it can be understood that establishing the relationship between passive leading indicators and activities is useful, pursuing the same approach for active leading indicators has more benefits.
The safety leading indicators have been evaluated in a location-based manner using BIM [22]. Near-miss detection tools and hazard identification tools have been another way to visualize the effects of the safety leading indicators [23]. However, a location-based visualization is not enough; the time gap between the recognition of each safety leading indicator and occurrence of the related safety hazard has an important role to take preventive measures. Moreover, some leading indicators even when occurr in a specific place of the project may affect another place that was not considered in previous works.
The effects of some of the specific types of safety leading indicators have been evaluated and their impacts have been validated [24]. Specific types of safety leading indicators have been studied before or the safety leading indicators related to specific types of activities [25]; this highlights the need for a comprehensive approach to assessing the influences of safety leading indicators on a project and evaluating their impacts on the projects in a way to recognize all of their effects on the project [26]. Moreover, monitoring and realizing the occurrence of safety leading indicators is another step to prevent the hazards from happening in a project [23]. Nonetheless, to fully assess the impact of safety leading indicators in a project, a safety leading indicator measurement tool is needed to not only monitor and detect the safety leading indicators but to predict the occurrence and disappearance of them.
To fulfill these gaps in this study; (1) the relationship between safety leading indicators and activities are considered by defining a set of attributes and assigning them to each safety leading indicator and activity and establishing the relationship between safety leading indicators and activities by checking whether they have matching attributes; (2) after detecting the safety leading indicators a feature is defined for the safety leading indicators which evaluates the effective period of safety leading indicators. Moreover, the impact of the safety leading indicators on places other than the locations in which the related activity is taking place, is considered by defining a term which is named the effective radius; (3) to assess all the safety leading indicators affecting the project, a knowledge base is built and the safety leading indicators are stored in it by utilizing different sources of knowledge; (4) to assess the safety leading indicators through project's timespan and considering the appearance and disappearance of leading indicators a real-time approach is utilized so that each safety leading indicators can be added or deleted using the safety leading indicator measurement tool.
This study aims to develop a safety leading indicator measurement tool considering the safety leading indicators. This system utilizes the BIM of the project and extracted safety leading indicators in the knowledge base to assess the impact of safety leading indicators in a time-and location-based manner. This system considers the relationship between safety leading indicators and activities of the project and determines that when and where each safety leading indicator is going to occur and have an impact. After calculating the cumulative impact of all safety leading indicators of the project, it generates a heat map through time to visualize their influence. This system determines that in each location and time how likely it is for a hazard to occur based on active safety leading indicators.
2 Literature Review

Safety Leading Indicators
Many studies have tried to develop safety predicting models to prevent hazards [27]. Moreover, the relationship between safety climate and safety behaviors was considered [28]. Many other factors that have safety predictive value also have been studied [29]. Near-miss detecting models that detect the events that had the potential to lead to injuries but they did not result in hazards [30], real-time location tracking systems which are based on tracking workers or equipment and alarming them proactively before any hazards occur [31], have been used as means to determine the hazards beforehand to prevent them.
Hierarchy of the controls is a reliable theory in the field of safety in construction projects [32]; it ranks metrics that are implemented addressing safety issues in a project over a systematic manner. A typical ranking includes elimination, substitution, engineering controls, administrative controls, and personal protective equipment (PPE). These controls can be divided into two parts; higher ranks (respectively, from the first to the third controls) which are mainly implemented in the early stages of project lifecycle, such as conceptual and detailed design, and lower ranks (the last two controls) which are mostly related to later stages, such as construction and operation. Although the theories such as the Szymberski curve suggest that the influence level of safety strategies is higher in the first ranking part, the second part should not be ignored and safety management systems should cover the total project lifecycle [33]. In other words, the safety management system's performance will be more efficient when assessments start at the early stages of a project and that it continues, while a project is progressing by monitoring the safety performance and using active safety leading indicators is a way to improve the safety performance of the projects from the early phases.
In general, the main purpose of using safety indicators in monitoring safety levels in a construction project is to provide information for managers to make decisions regarding how and where to perform an action; actions carried out based on safety indicators aim to avoid future accidents. Grabowski et al. stated that recognizing indicators before the occurrence of an accident potentially improve projects' safety [34].
Safety indicators in construction projects are divided into two groups; leading indicators as proactive or predictive indicators [35] and lagging indicators which are collected after the occurrence of accidents to assess consequences [36]. It has been found particularly over the last years that lagging indicators themselves do not provide enough information and insight to avoid future risks effectively and their required workability has been questioned [23,34,37,38]. On the other hand, leading indicators are highly reliable, because they enable managers to find deficiencies, and correct them to prevent or mitigate their future impacts [39]. Sinelnikov et al. mention that leading indicators are project-oriented and are focused on a site level, because they are process-specific and that construction projects differ in terms of their size, location, operations, organizational culture, etc. [12]. Wang et al. mention the following characteristics for safety leading indicators: specific, measurable, achievable, relevant, and time-bound [40]. A combination of this fact and theories, such as the hierarchy of controls, reveals that each safety management system relying on leading indicators should utilize previous information at the early stages of the project and evolve during later stages of project progress, while new leading indicators would appear.
Leading indicators can be classified as passive or active [41]. Passive leading indicators are often set in higher levels of the organization and content of strategies adopted before the beginning of a project. Active leading indicators are measured and adjusted dynamically both in pre-construction and construction phases allowing real-time implementation of risk mitigation practices [42].
Costin et al. classifies active safety leading indicators into three main branches: (1) equipment strucks which are produced by impacts between workers, objects and equipement; (2) slips, trips, and falls from height which are conditions that may lead to workers falling and injury and; (3) physical stress by overexertions which are conditions that cause injuries every year as they decrease workers movability in construction sites [43]. Shen et al. inumarates dangerous equipement and equipement proximity as the main origins of excavation related hazazrds [44]. Cheng et al. applied a real time location sensor to track worker movements in construction phase and determined that increasing in workers movement and conjestion leads to a higher probability of accidents occurring [45].

Information Technology and Safety Management Systems
Information technologies are addressing rapidly the managerial aspects of the construction industry [46]. For example BIM has been used to evaluate the progress of projects during their construction [47]. Visualization technologies, such as 3D/4D Computer-Aided Design (CAD), Virtual Design and Construction (VDC), and BIM, have been widely used to enhance safety performance in projects by facilitating safety communication and coordination between line managers and technical workers [48,49]. Moreover, database and automatic rule-based checking technologies are utilized in safety management systems [50]. Some scholars have developed add-ons based on platforms provided by existing BIM tools. Shen and Marks developed a framework for safety personnel to view nearmisses and their frequency to identify hazardous areas [51]. Yuan et al. also developed a tool to prevent the safety hazards in the design phase; it helps to automatically inspect safety risks based on existing safety regulations and documents [33]. Lim et al. integrated the WBS of the project and the BIM 4D model to reduce develop an effective safety plan for projects [52]. Information technology has also facilitated safety management in construction projects and has shown great potential in process automation and efficiency improvement [53,54].
Building information modeling has been integrated with knowledge management methods to utilize and store the information extracted from previous designs and projects [55]. Lu et al. have integrated safety risk data and building information modeling to propose a quantitative method to assess the safety risks involved in a construction project [56]. Moreover, Takim et al. have also developed a safety risk detection method that utilizes knowledge management methods and BIM to identify the risks that workers are exposed to in projects [57].

Knowledge Management for Safety Leading Indicators
Human knowledge must be imparted, to use in computer systems. For doing this, Maravilhas and Martins divided human knowledge into explicit and tacit ones [58]. Data gathered from previous projects and literature are considered explicit knowledge, and knowledge of construction site workers are considered implicit [59]. Kamara et al. concluded that adoption of implicit knowledge is difficult in construction projects due to their complexity and uniqueness [60]. Hence, converting the active leading indicators recognized by workers on-site to data that can be used in a safety management system is complicated. Acquiring, sharing, expressing, and delivering are steps toward converting implicit knowledge into explicit [60]. Knowledge management has a different usage in each level of the project's positions. Moreover, the level of access to information must be customized based on the hierarchy of the organization. For instance, type of the information that each construction manager, project manager, executive manager and, etc. in the organization needs, varies based on their position. Therefore, bulk information must be put into knowledge management process, so it can be used optimally [61]. In addition, although the need for construction projects for acquiring safety knowledge is tangible, knowledge storage and transfer systems are inefficient for both internal and external resources [62]. Therefore, an efficient knowledge management method needs to be developed for managing the implicit knowledge of on-site workers related to safety leading indicators. Applying implicit and explicit knowledge of leading indicators through BIM platforms makes the share and reuse process possible, based on a 3D model.

Overview of the Research Framework
The study here is aimed at developing a safety leading indicator measurement tool relying on safety active leading indicators by integrating BIM and knowledge base approaches. BIM model of construction project and knowledge base of past and present leading indicators formed the foundation of the proposed tool here. In addition, an add-on is developed to aid in integration and facilitate the process of information exchange between the main components of the proposed system. In the first step, a safety leading indicator knowledge base is built to encompass three parts: identification of active safety leading indicators, recognition of attributes related to leading indicators, and forming relationships between them. It is worth noting that in this research, data can be added to the knowledge base during the construction phase. Hence, classification of data collected from different perspectives is essential to reach the same format which increases the efficiency in process of using information. Then, the project's 3D model is generated inside the BIM environment, e.g., Autodesk Revit. This makes the information exchange, extension, and integration possible among different BIM-based software which is needed for developing the presented system. Afterward, project activities are identified. Furthermore, the model in Revit is exported to Navisworks. In the next step, activities, and their features and attributes are defined. Attributes indicate whether this activity is vulnerable to a certain event like a fire or heavyweight, or not. Moreover, the add-on extracts model elements and assigns their relative activities. In this step, the activities which affect each element through the period of the project, are assigned. In the fourth step, identified safety leading indicators, and their features and attributes are defined. Safety leading indicators' attributes indicate whether a safety leading indicator is capable of causing a certain event or not.
In the final step, the defined safety leading indicators are allocated to the activities. To achieve this goal the add-on checks whether the safety leading indicators and the activities have matching attributes, or not. In addition, the impact of safety leading indicators on each element is investigated in this step. As a result, a heat map is generated which visualizes the effects of the safety leading indicators on the project status in time and location. Through this heat map, critical times and locations are recognized. Thus, project staff can use this information to select strategies to prevent accidents. An overview of the research can be found in Fig. 1.

Knowledge Management
A bulk of information can be attained by exploiting best practices, documents, and experts' expertise. Microsoft Access is used in this study to utilize and manage the knowledge which was accumulated with the mentioned methods. To measure the effects of the safety leading indicators, this establishment is necessary. This knowledge database when implemented in the safety leading indicator measurement tool, can make the evaluation of the safety leading indicators' effect possible. Building the information database for safety leading indicators is described as follows: (1) Collecting the safety leading indicator knowledge.
As mentioned before, the experts' expertise, documents, and best practice provide ample safety knowledge. This knowledge of safety leading indicators presents details about the indicators influencing the project and their attributes. (2) Determining the knowledge elements for safety leading indicators. The work breakdown structure of the project is used to detect the under-progress elements and the activities assigned to them. Using attributes of the safety leading indicators and activities, they can be linked together.

Development of Automatic Inspection Tool Based on BIM
In the first step, a connection between schedule, safety leading indicators, and the 3D model is acquired. To utilize BIM capabilities, Visual Basic.net (VB.NET) as a programming language is employed to make the connection possible In Navisworks. Revit is used to develop a 3D model of the project and the results are then exported to Navisworks. Navisworks allows the use of Application Programming Interface (API) and the Microsoft Visual Studio to develop the safety leading indicator measurement tool. All the 3D details are included in the model; the time factor is also added to the model with an add-on.
All BIM graphical and parametric data is extracted with the add-on, and all the elements' boundaries are defined. Hence, the safety leading indicators are attached to their relative elements and their radius of effect is specified in the add-on.
By applying the impact of the safety leading indicators of each activity on their relative elements through time, two results are extracted: (1) a heat map which visualizes the effects of safety leading indicators in time and any location and (2) a textual result which indicates how safety leading indicators impact the project regarding time and space.

Establishment of a Safety Leading Indicator Knowledge Base
Attaining, storing, and utilizing the knowledge related to safety-leading indicators is the main purpose of creating a knowledge base. To carry out the operation discussed in Sect. 3.2, three processes (i.e., collect the safety leading indicators knowledge, establish the corresponding relationships, express and store the safety leading indicators knowledge) should be performed. The three processes are described as follows:

Collecting the Safety Leading Indicators Knowledge
The information regarding the activities and elements can be extracted from project characteristics. To extract the safety leading indicators, best practices, documents, and experts' expertise are utilized as the main sources of knowledge. Methods used to elicit this information are specified in the following paragraphs: (1) Experts' expertise: although this source of information is implicit, it is one of the paramount sources regarding safety leading indicators. The experience that experts have accumulated over time can be exploited. In this case, their opinions should be questioned and then converted to knowledge.

Establishing the Corresponding Relationships
The safety knowledge base includes 4 main parts: (1) activities (2) safety leading indicators (3) Their attributes (4) the corresponding relationships between them. Activities and safety leading indicators should be connected. To establish the relationship between them, their attributes are utilized. Activities and safety leading indicators possess attributes. The first step toward establishing the relationship between safety leading indicators and activities is to appoint their attributes. Activities' attributes refer to being vulnerable to a safety hazard. On the other hand, the attribute of the safety leading indicator represents the hazard that the safety leading indicator may cause. If a safety leading indicator and an activity have matching attributes then the correspondence accedes between them. Hence, indicators affecting each activity can be detected.

Expression and Storage of the Safety Leading Indicators Knowledge Base
To improve the utilization of the gathered data, it should be stored in a knowledge base. Safety leading indicators knowledge, activities, attributes, and their corresponding relationships when stored and expressed can form a knowledge base. The knowledge is categorized into four sections (i.e., activity, attributes, activity-element, safety leading indicator) and three relationships (activity-safety leading indicator, activity-attributes, safety leading indicator-attribute) that have the structure of ''IF…, THEN…''. The storage method is described as follows; to store the extracted knowledge a database management software is used. Microsoft Access has two types of storage forms: ''tables'' are used to store the information, and the other form is used to establish the relationships between these tables by merging the data stored in the tables.
(1) Structure design of knowledge storage tables. Four knowledge categories (i.e., activity, attribute, activity-element, safety leading indicator) are accumulated in storage tables. The features regarding each category are illustrated in Tables 1, 2, 3 and 4. (2) Designing the relationship of knowledge storage tables.
Usually, two types of relationships exist between knowledge tables of the safety leading indicators: many-tomany and one-to-many. A many-to-many is the relationship in which a wide variety of knowledge items in the first table are related to a wide variety of knowledge items in the second table. Furthermore, a wide variety of knowledge items in the second table are also connected to a wide variety of knowledge items in the first table. On the other hand, a one-to-many relationship is when one knowledge item in the first table is related to a wide variety of knowledge items in the second table, and a wide variety of knowledge items in the second table are connected to when one knowledge item in the first table. In this study, to demonstrate the relationship between knowledge storage tables, the many-to-many relationship is used. The manyto-many relation table is illustrated in Figs. 1 and 2.
To illustrate the relationship between two tables a third table must be used to express the many-to-many relationship between them. In this study, the three kinds of relationships that exist between the tables are; activity-safety leading indicator, activity-attribute, and safety leading indicator-attribute. These relationships must be stored in a relational table to show the connection between the activities, safety leading indicators, and attribute tables. Tables 5, 6 and 7 show the features of three relational data tables for the safety leading indicator knowledge base (Fig. 3).

BIM-Based Safety Leading Indicators Knowledge Base Application
Two kinds of tools can be developed to assess the impact of indicators; the first is to develop a tool that uses industry foundation classes (IFC). In this kind of tool, an exterior program is developed which utilizes IFC as input. Another approach is to develop an add-on based on BIM. Since using IFC as input has its disadvantages, such as missing data when used to transport information into the third program, developing an add-on can be more effective for operators.     Figure 4 illustrates the safety leading indicator knowledge base used in the developed application to distinguish the effects of the safety leading indicators on elements through time. To develop the BIM model Revit is used; then the model is transported to Navisworks, where all the project information is involved in the model according to Sect. 3.3. In this section, the development of the add-on is characterized.

The Algorithm for Measuring the Impact of the Safety Leading Indicators
The initial step in developing a program to measure the impact of safety-leading indicators is to design an algorithm fitting the purpose of the add-on. The proposed algorithm to develop the add-on based on Navisworks consists of three main parts, input, processing, and displaying the results which are described in detail in Sect. 5.2. In the input stage, the data including the database, model, the relationship between the model and database, the relationship in the database itself, and the settings data are given to the add-on. In the processing section, the data given in the previous stage is utilized to identify each day's safety leading indicators and also their relative elements. Finally, in this step, the cumulative impact of the indicators is calculated and stored. In the final stage, displaying the results, the calculated data in the

A Detailed Description of BIM Development
Three main parts of the algorithm are explained in this section in detail: (1) Input: in this stage developed BIM model, safety leading indicators knowledge base, activities affecting each element, attributes, safety leading indicators of each activity, and the settings data are given to the add-on. The setting input defines the boundaries of the heat map's color-changing range regarding the cumulative impact of each element's safety leading indicators in displaying the results stage. This provides the operator a way to define the regulations regarding the tolerability of safety leading indicators' impacts on elements according to the characteristics of the project. Furthermore, the compatibility threshold of the attributes is given to the program which indicates the minimum number of attributes to establish the relation between safety leading indicators and activities. Moreover, the attributes are added to the program in this section. Because time factors cannot be distinguished in Navisworks, cost, start date, and finish date of the activities are allocated using the project schedule. The established safety leading indicators knowledge base is explained in  [63]. Therefore, for the purpose of this research five-point Likert-scale is easier and more effective to utilize. The effective area has three types: the ones which have an impact on the assigned element, indicators that have an impact on the assigned element and the areas nearby the element with a given radius, and the ones which affect the whole project. The effective period of each safety leading indicator also has three types: from the date, indicator effect occurs through the start of the activity, from the start of the activity through the finish date of the activity, and from the finish date of the activity through the end of the project. Moreover, in this stage attributes and their factor of importance are given to the add-on, then, they are assigned to safety leading indicators and activities. To establish the relationship between safety leading indicators and activities two approaches are used: first, the relationship can be established manually, second, the program can assign them together using their attributes.
In the automatic approach, the program checks whether the safety leading indicators and activities have attribute compliance higher than the compatibility threshold using the important importance factor of each attribute, which multiplies the effect of the similar attribute when counting the sharing attributes between safety leading indicators and activities. The algorithm is shown in Fig. 5.
(2) Processing: to measure the impact of safety leading indicators on each element, the data given to the add-on in the input stage is used. In this stage, the first step is that the program goes through all the activities one by one. Then, the safety leading indicators assigned to each activity are identified. In this step, the data given to the program in the input stage is used to discover the area and period of safety leading indicators effectiveness. Then, the impact of each safety leading indicator on its relevant elements on each day is stored. Furthermore, the program checks whether it is the last activity or not. If yes, the program continues to the final step; otherwise, the program proceeds to the next activity. In the final step, the cumulative impact of every safety leading indicator of each element on each day is calculated. The algorithm is shown in Fig. 6.
(3) Displaying the results: in this stage, the impact of safety-leading indicators is displayed. To visualize the results, the heat map of the influenced elements is displayed day by day. The program starts to illustrate the heat map from the day which is given to the program, for a specific duration. On each day, the program assigns a color to each element. Three colors are used to distinguish the amount of impact influencing the elements. The yellow color is used to identify the elements that are impacted less than the lower limit given to the program in the settings. The orange color is used to identify the elements that are impacted between the higher and lower limit given to the program. The red color is used to identify the elements that are impacted more than the higher limit of the program. In the next step, the program checks whether it is the last element on that day or not. If yes, the program proceeds to the displaying stage; otherwise, the program continues to assign colors to the rest of the elements. After displaying that day's heat map the program checks whether it is the last day of the project or not. If it is, the program asks for another date to start displaying the results. If not, the program delays the displayed heat map, so each day's heat map can be distinguished from the next day's heat map, then proceeds to the next day. To apply this delay, multithreading is used to execute multiple parts of Navisworks at the same time. Multithreading prevents the program to malfunction which is derived from the stop command given to the Navisworks to demonstrate the delay. The algorithm for this step is illustrated in Fig. 7.

Case Study
A case study is introduced here to determine the effects of the safety leading indicators on a real project. A station in a water transfer line from Mamloo dam to a water refinery station located in the south of Tehran which is constructed in mountainous area, is modeled under Navisworks environment and safety leading indicators are extracted; see Fig. 8. As this project contained similar stations that some of them were already constructed, comparable safety records existed from constructed stations and the performance of the safety leading indicator measurement tool which was used in constructing the other stations could be evaluated using the existing records. Through this water transfer line, there are 35 other similar stations. 12 of these stations were already constructed before this safety leading indicator measurement tool was used. During the construction these stations 54 accidents were recorder which included falling from height, crashing with on-site equipment, crashing with falling objects, misfunctioning of dangerous equipment and etc. 13 of these accidents lead to injury and three of them lead to fatality. To evaluate the performance of the developed safety leading indicator measurement tool, it was used in the rest of the 23 unconstructed stations. To elaborate on how the safety leading indicator measurement tool functions, the information of utilization of this system on one of these stations is presented in this section. This station has a concrete structure and a 250 square meter area. Its construction was started on the 1st of November and finished on the 30th of December of 2020. Table 8 presents an overview of the case studies characteristics and the safety statistics of the 12 already built stations.
The first step is to build the knowledge base regarding the project specs. A database containing extracted safety leading indicators, activities, their attributes, activities affecting each element, and the relationships between safety leading indicators, activities, and attributes are established. To identify the safety leading indicators of this project, a literature review is conducted based on the project's nature which includes excavation, piping and concrete pouring. Adding the leading indicators that had been identified in constructed stations regarding the project's safety experts and documents, a long list is extracted which consisted the three main sources of knowledge which was mentioned in Sect. 4.1. After brain storming with project's staff a short list was created which can be found in Table 13.
As explained in Sect. 4.3, a database is created to store the safety leading indicators, activities involved in the project, and their attributes. To create the knowledge base, a list of safety-leading indicators should be extracted. The knowledge base is given in Tables 12, 13, 14 and 15. The method introduced in Sect. 4.2 should now be implemented in this step and the relationship between activities, attributes, and safety leading indicators should be established in a table in the knowledge base. Using this relationship, the impact of each activity in its associating area can be considered. The relational tables are presented in Tables 16, 17 and 18. After establishing the knowledge base, all the activities, attributes, settings, safety leading indicators, and their relationships are given to the program as input. The purpose of using the add-on is to discover the date and the elements that the impact of the safety leading indicators on them is critical. By utilizing the algorithm explained in Sect. 5, the cumulative impact of safety leading indicators on each element and each day is calculated and the results are displayed in a heat map, see Fig. 9.

Verification of the Developed Add-On
To validate this study, Yuan's approach is used [33]. To discover the effectiveness of the developed safety management tool a semi-structured questionnaire survey was sent to the associative of the project and researchers. Hallowell and Gambatese suggested a basis to select the participants of the questionnaire, which is used in this study [64].
Two parts were included in the questionnaire; the first part which is necessary for experts to have an understating of the study which contained all the details regarding the presented case-study, the background of the research, the implementation of the safety leading indicators knowledge base, and BIM, the tool developed to measure the impact of the safety leading indicators and how it is executed in the project. The second part captured the respondent's demographic information comprising of work experience, job title, age, educational status, and the most paramount part was to through their experience, rate the effectiveness of the developed tool using the five points Likert scale. 32 people responded to the questionnaire; their answers to the first section are shown in Table 9. It can be interpreted that all of the conditions mentioned by Hallowell and Gambatese are met to ensure that they have enough expertise in the field and, hence, the survey results are valid. A majority of the respondents (more than 65%) have 7 or more years of work experience and that all hold a bachelor's or a higher educational degree. Moreover, researchers scholars' field of study is the safety of the construction projects.
To ensure the effectiveness and efficiency of the developed safety leading indicator measurement tool 5 questions were responded by the participants. Extracted statistical results of the survey are presented in Table 10. The results from the survey indicate that the average rate of each of the five questions is above four meaning that compared to traditional methods, the developed safety leading indicator tool is more effective. It can also be inferred that the developed tool regarding the labor-saving aspect has a favorable function (Table 10).

Results and Discussion
According to the results from the add-on, the critical impact happens on the element of ID = (200614) from 1st December to 15th of December of 2020. After analyzing the critical dates, it is detected that the safety leading indicator of ID = (3) and the activity of ID = (1) have the most impact on the results. This information can be used by project participants to perform an action to prevent future accidents by changing the start date and finish date of some of the activities if it is possible or mitigating the safety leading indicators that have a critical impact on the project. The results of the utilization of this system in construction of the 23 unconstructed stations can be found in Table 11.
To compare the results, it must be mentioned that these stations are similar in size, construction method and the condition of the construction site. Before utilizing the safety leading indicator measurement tool, the number of accidents that occurred in constructing of the 12 of these stations was 54 and after that, 64 accidents happened in the rest of the 23 stations. After normalizing these numbers by calculating the overall accidents happening per station, its number decreased from 4.5 to 2.78. The number of injuries per station decreased from 1.08 to 0.56 and the fatalities per station decreased from 0.25 to 0 which indicates that the developed tool has an practical implication in construction projects.
Xu et al. have done a systematic review in 2021 about safety leading indicators and they have highlighted that safety leading indicators are measures to identify the weakness and strenghths of the safety management systems that are based on safety leading indicators. Moreover, they can determine situations that may lead to hazards and lead to proactive measures based on these situations [65]. In this study, critical safety leading indicators have been determined based on the projects charactrisitcs. These leading indicators can lead to hazards that actually had happened in the assessed project. Weakest locations and times have been determined using these indicators. Determining these indicators helped to develop proactive measures in the project. Moreover, using BIM visuailizes the effects of the safety leading indicators to facilitate the identifying of  Fig. 6 Algorithm for the process stage safety leading indicators that exist in the project and have crucial effect on the project. Regarding the theory of hierarchy of controls, the presents safety leading indicator measurement tool covers the both higher and lower ranks controls. The first application of this tool was in the early stages of the project, and its merit is that it can generate a overview of the project's safety condition which leads to enhancing the effectiveness of the adopeted safety strategies. The results of the utilizing the safety leading indicator measurement tool indicate that it is compatible with Szymberski curve. Moreover, the lower level ranks in the hierarchy of controls are also enhanced. The proactive manner in the presented tool ameliorates the administrative controls and workers safety while conducting the projects' activities.
As the developed system is based on proactive safety leading indicators, any safety leading indicator that is found during the project's construction phase must be added to this system by safety managers. Yet, having this process in the project management system had its own challenges, such as lack of sufficient perception of unconstructed parts of the project; therefore, it is recommended to add other tools such as virtual reality to this system to facilitate extracting the safety leading indicators during the project and adding them to the safety leading indicator measurement tool.

Conclusions
In this study, an add-on was developed to assess the impact of safety leading indicators on construction projects by integrating a safety leading indicator knowledge base and BIM; this allows the safety managers and project participants to make important decisions to prevent accidents. Safety leading indicators knowledge base was created by extracting information from expert's expertise, documents, and best practice. Furthermore, to develop the add-on, Navisworks API and Visual Studio's framework were used to determine the impact of safety-leading indicators. To verify the developed tool, a case study was considered, which illustrated that it is effective and functional to assess the impact of the safety leading indicators using the knowledge base. These efforts can provide safety managers   Fig. 9 Heat-map generated by the program and other project participants with a view of further incidents. Hence, they will be able to take proactive measures to prevent accidents. Finally, in the conducted survey the developed tool's effectiveness and efficiency were examined and confirmed. The results of this study could be put in use by project managers, safety managers, designers, and site managers; in the project level. A database could be developed utilizing extracted information from each project that the safety leading indicator measurement tool was used in, and in the early stages of a new project, this databse could be utilized to enhance the performance of the developed tool. The limitation of this study is that the knowledge base should be more comprehensive. Thus, to have a better view of the project, other resources should be studied. Furthermore, some parts of this study still rely on manual efforts, as to recognize the effective period of the safety leading indicators, it should be given to the program manually. Moreover, although this research considers the relation between safety leading indicators and activities, their attributes were assumed to be Boolean; each activity or safety leading indicator has the attribute or does not, other types of attributes should be considered to have more accurate results. Suggested future studies are: expanding the safety leading indicators knowledge base, using machine learning to facilitate the identifying of safety leading indicators effective period, and investigating other types of attributes of safety leading indicators and activities.
Funding Open Access funding enabled and organized by CAUL and its Member Institutions.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/. See Tables 12, 13, 14, 15, 16, 17, and 18.

15-Nov-20
Data Availability Statement All data, models, and code generated or used during the study appear in the submitted article.