Ontology for design of active fall protection systems

AUTCON-02220; No of Pages 16 Automation in Construction xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Automation in Construction journal homepage: www.elsevier.com/locate/autcon

Ontology for design of active fall protection systems Brian H.W. Guo ⁎, Yang Miang Goh Safety and Resilience Research Unit (SaRRU), Dept. of Building, School of Design and Environment, National Univ. of Singapore, 4 Architecture Dr., Singapore 117566, Singapore

a r t i c l e

i n f o

Article history: Received 28 July 2016 Received in revised form 24 January 2017 Accepted 28 February 2017 Available online xxxx Keywords: Ontology Knowledge engineering Construction safety Fall from heights Active fall protection systems

a b s t r a c t This paper aims to develop an ontology (AFPS-Onto) which formalizes the knowledge of active fall protection system (AFPS) design, with attempt to facilitate knowledge sharing and reuse. METHONTOLOGY was adopted as a method to build the AFPS-Onto. The AFPS-Onto consists of nine core concepts: hazard, actor, task, ifc building element, construction method, constraint, safety resource, hazard control measure, and residual risk. The concepts, relations, attributes, and axioms were coded using Protégé. The ontology was evaluated through automated consistency checking, criteria-based and task-based evaluation. The AFPS-Onto fills the knowledge gap by providing a formal and shared vocabulary for the domain of AFPS design. This can promote knowledge reuse and sharing among professional engineers. In addition, the ontology can be used to develop knowledge-based systems to help design effective AFPS. Future effort can be made to develop ontologies of other control measures against fall from heights and combine them into a fall from heights ontology (FFH-Onto). © 2017 Elsevier B.V. All rights reserved.

1. Introduction Over the lifespan of a building, in particular during the construction phase, workers are required to work at heights [1]. As a consequence, managing fall from heights (FFH) constitutes an essential part of construction safety management. In fact, FFH is a leading direct cause of fatalities in the construction industry across many countries, such as the US [2], the UK [3], Australia [4], and New Zealand [5]. In Singapore, falls accounted for 35% of all workplace fatalities in 2015 and more than half of the falls were contributed by the construction industry. FFH also accounted for around 20% of the major injuries in Singapore workplaces over the past five years [6]. Preventing FFH is a significant concern for different stakeholders in the construction industry, including government agencies, clients/developers, contractors, health and safety professionals, and workers [7, 8]. There is a hierarchy of control measures for FFH, ranging from elimination (e.g., prefabricating wall frames horizontally before standing them up), substitution (e.g., using mobile elevated work platform instead of ladders), engineering controls (e.g., guardrails), administrative controls (e.g., working-at-height rules and procedures), to personal protection equipment (PPE). Although the hierarchy considers PPE as the least effective control measure, it is a must in situations where working conditions are difficult and other controls are not applicable. PPE for working-at-heights includes active fall protection system (AFPS), which is “a means of providing fall protection that requires workers to take specific actions, including wearing (and otherwise using) personal ⁎ Corresponding author. E-mail addresses: [email protected] (B.H.W. Guo), [email protected] (Y.M. Goh).

fall-protection equipment and following prescribed procedures” [9]. It includes fall-arrest and travel restraint systems. A fall arrest system is designed to absorb the energy created by its user(s) during an accidental fall from heights. Typically, it consists of the following components: (1) full body harness, (2) connectors, (3) lanyard, (4) energy absorber, and (5) anchor [10]. A travel restraint system is a system that prevents its users from reaching an unprotected edge or opening [11]. In order for an AFPS to protect workers working at heights, an effective design is a prerequisite. Standards were developed to guide qualified persons designing effective fall arrest and travel restraint systems, such as Z359.6 [12], Z259.16-15 [11], and SS 607 [9]. However, inadequate designs of AFPS are still common. For example, Goh and Wang [13] evaluated eleven horizontal lifeline system (HLLS) designs in Singapore and found that none of the eleven designs was adequately endorsed or calculated. In addition, Hoe et al. [14] pointed out that current designs of AFPS by professional engineers (PEs) did not cover a wide range of critical areas and that some of design cases were not even accompanied by any calculations. In many of the designs, important factors were ignored such as dynamic forces created during a fall, the mobility needs of the workers, and safe access and egress. Poor designs of AFPS provide a false sense of security; injuries and fatalities could be caused when workers wrongly assume that they are under protection. Inadequate designs can be in part attributed to a lack of knowledge (e.g., calculation methods). In practice, PEs tend to use different terms, jargon, and vocabularies in their designs, which makes knowledge reuse and sharing difficult. In addition, as in the case of other engineering designs [15], PEs can benefit from using knowledge-based systems (e.g., rule-based expert systems or probabilistic expert systems using a

http://dx.doi.org/10.1016/j.autcon.2017.02.009 0926-5805/© 2017 Elsevier B.V. All rights reserved.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

2

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

Bayesian network) when they design AFPS. These systems can offer recommendations and solve problems by using a rich body of knowledge in a domain of interest [16]. Developing an ontology is often considered an important starting point to construct knowledge bases for these systems [17]. Unfortunately, to the authors' best knowledge, no ontology has been developed yet to represent the knowledge of the domain of AFPS design. This often results in inefficiencies and inconsistencies in the design process of AFPS. Knowledge reuse and sharing are also significantly hampered. Thus, this paper addresses the research gap by developing and evaluating an ontology (AFPS-Onto) for the domain of AFPS design in the building and construction industry. The ontology is aimed at providing a formal and explicit specification of a shared conceptualization of AFPS design domain. The ontology represents the domain knowledge of AFPS design and provides a computer understandable vocabulary for knowledge reuse and sharing and intelligent system development. The rest of this paper is organized as follows: Section 2 presents a literature review of fundamental concepts of ontology, safety ontologies developed in the construction industry, and design of AFPS. Methodology used to develop the ontology is described in Section 3. In Section 4, we present the AFPS-Onto in terms of its generic ontological model, concepts, semantic relations, attributes, axioms, and coding. Evaluation of the ontology is described in Section 5. In Sections 6 and 7 we present conclusions and limitations and future work, respectively.

Table 1 Examples of ontology development methodology.

2. Literature review

intelligence research, they are often developed for information integration, information retrieval, and expert systems [24,29]. Ontologies are also used as an important foundation for expert systems in which implicit knowledge from the axioms in an ontology can be derived for automated reasoning and solving different problems. In addition, ontologies are utilized to construct Bayesian networks for developing probabilistic expert systems [35–37]. As a computational artefact, ontologies provide computer systems with a computational framework of a particular domain [24]. By representing the domain knowledge in a machine-interpretable format, ontologies were developed and used by automated reasoning techniques to draw conclusions and solutions for different purposes in the construction industry [38–43].

2.1. The concept of ontology The concept of “ontology” has its roots in philosophy, where it is concerned with the nature of being and existence. In the past decades, it became a popular term inside a number of artificial intelligence (AI) communities, including knowledge engineering, natural language processing, and knowledge representation [18]. There are many definitions about what an ontology is and these definitions have evolved over the time [19]. Gruber [20] provided a popular one: an ontology is an ‘explicit specification of a conceptualization’ (p. 908). The conceptualization represents a specific world view on the domain of interest [21] and it is composed of concepts, attributes and relations between concepts. Neches [22] provided a descriptive definition which defines an ontology as “the basic terms and relations comprising the vocabulary of a topic area as well as the rules for combining terms and relations to define extensions to the vocabulary.” An ontology, as a representation vocabulary, describes the domain knowledge in such a way that the specification can be interpreted by computer systems [23,24]. It captures an agreement on a domain conceptualization among stakeholders in the domain [25]. An ontology defines a shared vocabulary in a coherent and consistent manner with which queries and assertions are exchanged among people in a specific domain [20]. A number of methodologies for ontology building have been developed by researchers since the early 1990s. In 1990, Lenat and Guha [26] explored how to represent common-sense knowledge of the world for the Cyc project. In later years, a number of methodologies were proposed, as presented in Table 1. The British Standards Institution published an ISO standard to develop ontologies in 2015. The standard specifies the framework and rules for developing ontologies in the Web Ontology Language (WOL) [27,28]. A detailed review of all methodologies is beyond the scope of this paper. A systematic review of methodologies for ontology building can be seen in [29,19]. Ontologies have been developed in many fields (e.g., knowledge management, computer science, and artificial intelligence) for various purposes [24]. For example, in the domain of World-Wide Web, ontologies were developed to categorize websites and products for search engines (e.g., Google) and online shops (e.g., Amazon). Ontologies are used for knowledge connectivity, knowledge abstraction, and automation in knowledge processing in computer science. In artificial

Methodology

Ontology development process

Grüninger and Fox's approach used in TOVE project [30]

(1) Identify motivating scenario; (2) define informal competency questions; (3) define the terminology of the ontology; (4) define formal competency questions; (5) specify the definitions and constraints on the terminology; (6) test the competency of the ontology (1) Identify purpose; (2) build the ontology (ontology capture, ontology coding, and integrating existing ontologies); (3) evaluation; (4) documentation; (5) guidelines for each phase (1) Specification; (2) knowledge acquisition; (3) conceptualization; (4) integration; (5) implementation; (6) evaluation; (7) documentation (1) Determine the domain and scope of the ontology; (2) consider reusing existing ontologies; (3) enumerate important terms in the ontology; (4) define the classes and the class hierarchy; (5) define the properties of classes-slots; (6) define the facets of the slots; (7) create instances (1) Kick-off; (2) refinement; (3) evaluation; (4) ontology maintenance

A skeletal approach [21,31]

METHONTOLOGY [32]

a Simple Knowledge-Engineering Methodology (SKEM) [33]

The On-To-Knowledge methodology [34]

2.2. Safety-related ontologies In recent years, the roles played by ontologies in safety knowledge sharing and dissemination have been emphasized due to its ability to alleviate the interoperability problem in knowledge sharing and dissemination [44]. A number of safety ontologies were developed in the construction industry to formalize different types of domain knowledge and server for different specific purposes. The fundamental idea is that there is much to gain if safety data, information, and knowledge can be formalized based on a common set of ontologies that facilitates interoperability and reasoning process and improves efficiency of construction safety management. For example, Le et al. [45] developed a social network system for sharing construction safety & health knowledge (SNSS). The SNSS was aimed for better communication and representation for construction safety knowledge using a semantic wiki web and ontology approach. In addition, ontologies were developed to facilitate automated safety management, particularly job hazard analysis and management. For example, Wang and Boukamp [46] developed an ontology-based representation and reasoning framework for supporting job hazard analysis (JHA). The ontology represents and structures the knowledge about construction activities, job steps, and hazards. It forms a foundation for reasoning process which facilitates identification of potential solutions for hazards. In addition, Chi et al. [47] developed and used a construction safety domain ontology to match safe approaches identified in existing resources with unsafe scenarios. It aimed to reduce the level of human effort required in JHA and enrich the solution space by serving as initial references. More recently, Zhang et al. [48] proposed a construction safety ontology to formalize

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

job hazard analysis (JHA) knowledge. The ontology consists of three ontological models: construction product model, construction process model, and construction safety model. Based on the three models, job hazards and management solutions are linked with construction process and product. Similarly, Lu et al. [49] developed a Construction Safety Checking Ontology (CSCOntology) to represent knowledge used in safety checking in the construction industry. The CSCOntology integrates concepts like “line of work”, “task”, “hazard”, and “solution”, and semantic relationships between them. The ontology was utilized as the foundation of an automated construction safety checking system. 2.3. Design of AFPS Fall from heights can be caused by a wide range of organizational, human, and engineering factors. In a review of factors influencing the risk of falls in the construction industry, Hu [50] found that the use of personal protective equipment (PPE) and methods is one of strong casual factors. This factor covers two possible scenarios: workers' misuse/non-use of PPE and inappropriate design of PPE such as AFPS. Conventionally, significant attention has been paid on motivating workers to use PPE when working at heights [51,52], with less emphasis placed on the second scenario. In order to support and facilitate the design of AFPS, standards were developed at the national level, such as ANSI/ASSE Z359.6:2009 by American National Standards Institute, American Society for Safety Engineers [12], Z259.16-15 by Canadian Standards Association [11], SS 528 (specifications for personal fall-arrest systems), SS 541 (specifications for restraint belts), SS 570 (specification for personal protective equipment for protection against falls from a height. Single point anchor devices and flexible horizontal lifeline systems). Despite these standards, the design of AFPS has been problematic in the construction industry. For example, Goh and Wang [13] evaluated eleven cases of HLLS design and found that PEs adopted different assumptions (most are wrong), focused on different aspects, and produced design reports in different structure and formats. They concluded that all eleven cases were inadequate and that many of the PEs did not understand how a HLLS is supposed to work. In addition, the knowledge level of calculation methods is generally low [14]. PEs tend to focus on calculating static force rather than dynamic force on lifeline and users. Determining which type of AFPS should be used is not a straightforward process; it is dependent upon complex and dynamic construction environments and job tasks. Designers, especially novices, can benefit from decision support systems using case-based reasoning, rule-based reasoning, or Bayesian network. Note that the core of any decision support systems is knowledge [53]. Haghighi et al. [53] suggested that better knowledge management for decision support can benefit from

3

formalization of knowledge of the domain of interest using domain ontologies. Unfortunately, there is no ontology providing a structured and formal representation of knowledge in the domain of AFPS design, which makes the development of such decision support systems and knowledge reuse and sharing difficult. 3. Methodology This study adopts the METHONTOLOGY as a method to developing the AFPS-Onto. METHONTOLOGY was created in the Artificial Intelligence Lab from the Technical University of Madrid [32]. It consists of seven main steps: specification, knowledge acquisition, conceptualization, integration, implementation, evaluation, and documentation (see Fig. 1). METHONTOLOGY was adopted in this study based on the following considerations. First, it is a methodology for building ontologies either from scratch, reusing other ontologies as they are, or by a process of re-engineering them [19]. This enables the authors to develop the AFPS-Onto from scratch and, at the same time, integrate other ontologies (e.g., ifc building element). Second, unlike the Grüninger and Fox's approach [30] and the On-To-Knowledge methodology [34], METHONTOLOGY is an application-independent, as the ontology building is independent of the uses of ontology. This means that ontologies developed by using the METHONTOLOGY can possibly be used in various applications. With respect to maturity, METHONTOLOGY is considered the most mature approach to developing ontologies [19]. Specification focuses on determining the purpose, scope, level of formality, intended uses, and end-users of the AFPS-Onto. First, a set of competency questions were established in this step. These questions include, but are not limited to, “What are the purposes of the ontology?”, “Who are its end-users?”, “What information should be captured in the ontology?”, and “What design criteria should be followed?”. By answering all competency questions, this step produced an “ontology requirements specification”, which severed as guidelines for the whole ontology building process. Ideally, a global domain ontology, construction safety hazard management ontology (CSHM-Onto), can be developed to facilitate the design of control measures for various safety hazards in the construction industry. To develop such a global ontology, it is beneficial to adopt a bottom-up approach by developing local ontologies. These local ontologies represent knowledge of different hazard controls in the hierarchy that includes elimination, substitution, engineering, administration, and personal protective equipment. An advantage of the bottom-up approach is that it is more detailed than a top-down approach and the local ontology remains useful as it is developed. Local ontologies can be integrated and mapped into an intermediate-level ontology that focuses on a specific hazard (e.g., fall from heights). Fig. 2 illustrates the architecture of the CSHM-Onto, in which

Fig. 1. Development process of AFPS-Onto.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

4

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

a set of hazard-oriented ontologies are located at the intermediate-level and ontologies of different control measures at the local level. Each intermediate-level ontology can be seen as a module which shares the ontological model to allow interoperability between other modules. The interoperability can facilitate future ontology integration and mapping. From a local perceptive, in contrast, each ontology can be used as an independent representation of domain knowledge. The ontology (i.e., AFPS-Onto) developed in this paper represents an effort to build a local ontology within the hierarchy of control measures of FFH. Its main purpose is to formally represent the domain knowledge of active fall protection system design. The AFPS-Onto is an important part of the declarative knowledge, containing the basic concepts of the active fall protection system design and the relationships among them. It does not cover the design of positioning systems and passive fall-protection systems such as guardrails and nets. The ontology is expected to express the viewpoints and satisfy the informational needs of multiple stakeholders, including legislators, PEs, and health and safety practitioners. In order to acquire knowledge for the ontology development, six semi-structured interviews were conducted with experts, including one health and safety manager, one general manager of a fall protection systems supplier, a deputy general manager of a contractor, two PEs, and one fall protection system expert. Participants had rich experience in the field of AFPS design and FFH hazard management. The position and experience of all interviewees are listed in Table 2. The purpose of interviews is to elicit the knowledge of these experts and identify key concepts, attributes, and their relationships within the domain. Key interview questions include: (1) Can you describe different types of AFPS? (2) Can you describe the process of designing an AFPS? (3) What attributes of users (e.g., workers) do you consider in the design of AFPS? (4) What attributes of a task performed by workers are important when you design an AFPS? (5) How do you choose an appropriate anchorage point? (6) When you design an AFPS, is it important to know the

Table 2 Position and experience of interviewees. Interviewees Position

Industry experience

1 2

15 years 13 years

3 4 5 6

Health and safety manager General manager of a fall protection systems supplier Deputy general manager of a contractor PE PE Fall protection system expert

12 years 10 years 23 years 10 years of experience as practitioner in working-at-heights; 7 years of experience in delivering working-at-heights training

surrounding environment of a working-at-height task? (7) What are typical working-at-height scenarios to which we should apply a horizontal lifeline system, vertical lifeline system, or travel restraint system? (8) As a safety equipment supplier, how do you collaborate with contractors and professional designers/engineers? (9) Why sometimes professional designers/engineers develop inappropriate fall protection system? (10) What factors are important to consider when you design an AFPS? Results of the interviews are presented in Table 3. Concepts, attributes, and relations derived from interview partly capture the knowledge of AFPS and therefore they are included in the AFPS-ontology. Considering that it was not possible to identify all relevant concepts, attributes, and relations by interviews, a number of AFPS design standards were studied to complement the knowledge gained from interviews. These standards include: Specification for design of active fallprotection systems SS 607 by Singapore Standards Council [9], ANSI/ ASSE Z359.6:2009 by American National Standards Institute and American Society for Safety Engineers [12] and Z259.16-15 by Canadian

Fig. 2. Architecture of CSHM-Onto.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx Table 3 Summary of interview results. Key aspects

Results

Concept

User, anchorage point; task; work platform; surrounding structure; height Attributes User: weight; number; work posture; movement direction Task: frequency; duration Work platform: slope; shape Anchorage: accessibility; ease of connecting; shape Relations Relations among different active fall protection systems; relations among lifeline, connecting device, anchorage, and active fall protection systems;

Standards Association [11]. They specify requirements for the design and performance of complete active fall protection systems, including travel-restraint and vertical and horizontal fall-arrest systems. The standards represent structured expert knowledge and industry norms in the domain. During the process, key concepts of PPE were identified, including connecting device, body holding device, energy absorber, anchorage subsystem, and types of active fall protection system. In addition, eight environmental constraints that are important to the design of AFPS were also identified, including abrasion, chemical, cut, extreme weather, moving machinery, fire, heat, and welding flash. Moreover, forty-one AFPS design cases were reviewed to closely examine the relationships among concepts identified from interviews and design standards. These design cases reflect how designers communicate the knowledge of AFPS design with their clients, safety equipment suppliers, contractors, and the construction industry as a whole. Thus, they to some extent represent the formal language of AFPS design used within the community. By studying these cases, a number of important attributes were derived, including anchorage shape, work platform shape, relative position between anchorage and work platform, and leading edge of work platform. These attributes are included because they are discriminatory and of high relevance when choosing an appropriate solution to a working-at-height problem (e.g., dismantle scaffold). This can improve the usefulness of the AFPS-Onto for the future development of knowledge-based systems. In order to build the concept taxonomy, this paper adopted a middle-out approach, namely “identifying first the core of basic terms, then specifying and generalizing them as required.” [21,54]. For each

5

concept, the following intermediate representations (IR) were built: (1) a concept dictionary containing all the domain concepts and instances of such concepts, (2) a table of binary relations, (3) an attribute table for each concept and instance in the concept dictionary, and (4) a logical axioms table. An IR refers to the data structure used internally by a compiler to represent source code. The concepts, relations, and axioms were then coded and modelled by using Protégé 5.0. The ontology was evaluated through automated consistency checking, a criteria-based and a task-based evaluation method. Automated consistency checking was conducted by using the Description Logic reasoner: Pellet, a third-party plug-in in Protégé. The purpose was to detect unsatisfactory concepts (or inconsistent ontologies) and support the diagnosis and resolution of possible bugs. Four core criteria were selected from evaluation criteria proposed by past studies [20,55, 56], including clarity, extendibility, completeness, and coverage. These core criteria were used to evaluate and improve the AFPS-Onto through its whole development process.

4. AFPS-ontology 4.1. A generic ontological model In order to facilitate the future integration and mapping among ontologies at the same level, it is important that these ontologies are developed based on the same framework. Using a generic ontological model ensures that the vocabularies that are used in ontologies at intermediate and local levels do not conflict so that the interoperation of components is not problematic. Fig. 3 presents the generic ontological model which acts as a harmonization framework for developing local ontologies. The ontological model consists of three main parts: problem (i.e. specific hazards that need to be managed), context (i.e. the situation in which the problem exists), and solution (i.e. hazard controls). The problem includes five concepts: hazard, actor, task, Industry Foundation Class (IFC) building element, and construction method. These five concepts are used together to provide detailed information about specific safety problems. Context is composed of two concepts: constraint and safety resource, which are aimed at providing information about the context in which safety problems exist and solutions are designed. Solution is composed of two concepts: hazard control measure and residual risk. From a global perspective, the three parts are interconnected: a

Fig. 3. A generic ontological model.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

6

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

Fig. 4. Taxonomy of constraint.

Fig. 5. Taxonomy of hazard control measure.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

Fig. 6. Key attributes of main concepts.

Fig. 7. Screenshot of the ontology editor Protégé.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

7

8

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

problem is solved by solutions which, in turn, are constrained by the context in which the problem occurs. 4.2. Concepts 4.2.1. Task Task is used to represent the hierarchy of construction process. According to the definition provided by IFC [57], a task is an identifiable unit of work to be carried out in a construction project. Types of tasks include: attendance, construction, demolition, dismantle, disposal, installation, logistic, maintenance, move, operation, removal, and renovation. Each task consists of different types of sub-tasks. For example, “laying the formwork boards on the metal structure of shores and steel beams” is an instance of the Task “construction”. 4.2.2. Actor Actor defines all actors or human agents involved in hazard management. From the perspective of hazard management, the AFPS-Onto focuses on and defines the following actor roles: worker and professional engineer. A worker is defined as “any person who is protected from falling by an active fall-protection system, or in the case of a fall-arrest system, any person who might fall while attached to the system.” A professional engineer is defined as “a person who holds an engineering license or temporary engineering license in the province or other jurisdiction in which he or she is applying a design standard.” [9]. 4.2.3. IFC building element IFC building element comprises all elements that are primarily part of the construction of a building [57]. The AFPS-Onto integrates the structure of IfcBuildingElement from IFC schema. The class Ifc Building element includes the following subclasses: IfcBeam, IfcBuildingElementProxy, IfcChimney, IfcColumn, IfcCovering, IfcCurtainWall, IfcDoor, IfcFooting, IfcMember, IfcPile, IfcPlate, IfcRailing, IfcRamp, IfcRampFlight, IfcRoof, IfcShadingDevice, IfcSlab, IfcStair, IfcStairFlight, IfcWall, and IfcWindow. 4.2.4. Construction method Construction method represents methods, techniques, and technologies that are used to construct a building. Instead of relying only on PPE, construction methods can be improved to eliminate or minimize the effects of hazards. Innovative methods include off-site construction, prefabrication, preassembly, and modularization [58–60]. Off-site construction provides a wide range of benefits with respect to time, quality, cost, productivity, and safety [61]. For example, a Cross Laminated Timber (CLT) and Glued Laminated Timber (Glulam) method was adopted in Singapore to build beams, columns and long-span roof structure of a sports hall project [62]. Compared to conventional construction, the method saves manpower and reduces risk of fall from height. Other similar technologies that can minimize fall from heights include prefabricated prefinished volumetric construction (PPVC) [63]. Incorporating construction methods within the ontology allows the extendibility of the ontology to integrate the concept of design for safety (DfS) [64]. For example, roof safety wire mesh is a widely used DfS idea for improving safety in roof installation and maintenance. An integration of DfS ideas into the ontology can facilitate the development of ontology-based expert systems for safety. 4.2.5. Hazard A hazard is anything in the workplace that has potential to harm people. Common safety hazards in the construction industry include: struck by moving objects, fall from heights, caught-in/between objects, and electrocution. As discussed earlier, the AFPS-Onto focuses on fall from heights.

4.2.6. Constraint A constraint in the AFPS-Onto is something that restricts and/or regulates Construction method and Hazard control measure. The concept of Constraint is in part adopted from El-Gohary and El-Diraby's IC-PROOnto model [25]. With reference to Fig. 4, it consists of four main types: environmental constraint, project constraint, standards, and regulatory constraint. In Fig. 4, ‘is-a relationship’ is used to represent the relationship between classes, whereas ‘instance-of relationship’ describes the relationship between classes and instances. • Environmental constraints represent environmental conditions that need to be considered when designing hazard controls. For example, all components of an active fall protection system should be selected by considering environmental conditions such as corrosion, chemical attack, heat, etc. [11] (p. 10). • Project constraints are limiting factors that are related to the construction project, including budget, schedule, and resource availability [25]. These limitations and restrictions can influence how hazards are identified and how solutions are designed and implemented in practice. • Standards represent any standards and codes of practice that are relevant to the design of AFPS. For the AFPS-Onto, these include, but are not limited to, ANSI/ASSE Z359.6:2009 by American National Standards Institute, American Society for Safety Engineers [12] and Z259.16-15 by Canadian Standards Association [11], and SS 528 (specifications for personal fall-arrest systems), SS 541 (specifications for restraint belts), SS 570 (specification for personal protective equipment for protection against falls from a height. Single point anchor devices and flexible horizontal lifeline systems). • Regulatory constraints represent constrains that specify relevant laws or regulations that must not be violated. Relevant regulations from Singapore include Workplace Safety and Health Act [65] and Workplace Safety and Health (Work at Heights) Regulations 2014 [66]. Table 4 A summary of ontology evaluation approaches. Approaches

Description

Gold standard This approach compares an evaluation ontology with another benchmark ontology by empirically measuring similarities between ontologies both lexically and conceptually [71]. Data driven This approach compares an evaluation ontology with a corpus by either performing automated term extraction on the corpus or counting the number of terms that overlap between the ontology and the corpus [72]. This approach uses Description Automated consistency Logic reasoner to assess the consistency of an ontology. checking Popular reasoners include Pellet, FaCT++, and Ontop. Criteria-based This approach utilizes a set of evaluation predefined criteria for evaluating an ontology [53]. Common criteria include: clarity, consistency, expandability, minimal ontological commitments, conciseness, completeness, coverage, and correctness. Task-based This approach evaluates an evaluation ontology by using it in tasks and assessing the performance. It is an effective approach to assess the capability of an ontology to achieve its purposes and objectives [53,71].

Limitations • Access to the “gold” ontology is not possible; • Lack of the gold standard; • Flaws in the comparison methods

• Not suitable for evaluating the correctness, clarity, and usefulness of an ontology

• Only checks internal consistency of an ontology

• Some criteria (e.g., correctness) lack quantitative and object measures; • To a large extent, rely on expert judgement

• Does not assess the structure, architecture, and design of an ontology

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

9

Onto is limited to the design of AFPS and thus focuses only on PPE. As shown in Fig. 5, the concepts of PPE include connecting device, body wear, anchorage subsystem, energy absorber, and active fall protection system. 4.2.9. Residual risk Residual risk refers to hazards and risks remaining after the inherent risks have been reduced by hazard controls. For example, even though an AFPS has been well designed, there still exist a number of residual risks such as “damage by sharp edges”, “improper fitting of PPE”, “poor installation”, and “inadequate training and supervision”. By including residual risk, the AFPS-Onto is able to provide a more complete picture of hazards and address limitations of adopted hazard controls. 4.3. Semantic relations

Fig. 8. Working-at-heights scenario.

4.2.7. Safety resource Hazard control measures require and consume safety resources so as to be effective. In the domain of construction safety hazard management, they are composed of the following: • Financial resource refers to the money available to operate hazard management. • Human resource includes “workers”, “safety professionals”, “project managers”, “engineers”, and “clients”. • Physical resource represents any tangible resources that can be utilized in hazard management, including equipment (e.g., existing fall protection systems) and existing anchorage. • Safety knowledge refers to knowledge required in hazard management. • Safety knowledge item refers to “physical or symbolic manifestation of” safety knowledge [25] (p. 733), including fall protection plan, safety instructions, and rescue procedure. 4.2.8. Hazard control measure Hazard control measure consists of a hierarchy of control: elimination, substitution, engineering controls, administrative controls, and personal protective equipment (PPE). Elimination refers to the total removal of a hazard, while substitution focuses on replacing the hazard by one that poses lower risk. Engineering controls are the use of machines, tools, structure and equipment to reduce the exposure to a hazard and likelihood of harm. In contrast, administrative controls are concerned with rules, procedures, and training programs that help reduce the possibility of harm caused by the hazard. Using PPE is considered a last resort or an additional protective measure, which are still frequently required in the construction industry. As discussed earlier, the AFPS-

Semantic relations are defined as “meaningful associations between two or more concepts, entities or sets of entities” [67]. Like concepts, semantic relations play a critical role in knowledge representation. Interconcepts relationships are an integral part of the design of the AFPSOnto. A relationship models an association between two or more concepts. An inter-concept relationship is typically defined by a verb phrase that describes the semantics of the relationship. The number of concepts involved in a relationship represents the degree of the relationship [68]. A relationship is context-dependent and therefore it should derive its meaning from being embedded in the context and domain in which the relationship is being constructed [25,68]. For example, the appropriate interpretation of “controls” (i.e., Hazard Control Measure controls Hazard) depends on the domain and context in which it is used. To summarize, the sematic relations used in the AFPS-Onto are categorized into the following types: • Hyperonym-hyponym (supertype-subtype) relations Hyponym refers to the narrower concept (e.g. Worker), and hyperonym is the broader concept (Actor). The hyponymy relation is the most fundamental semantic relation [67]. The relation includes the following variations [69]: “is_equivalent_to”, “is_similar_to”, “is_disjoint”, and “is_opposite”. An example in the AFPS-Onto is: “Rigid Rail System bis_disjointN with a Vertical Lifeline System”. • Meronym-Holonym b part-whole relationN relations This relation refers to the relation between a concept and its constituent parts. For example, in the AFPS-Onto, “Body holding device” b is_subsystem_ofN “Active fall protection system”. Its reverse relation is b has_subsystemN. • Concept-object (instance-of) relations The relation refers to association between types (classes) and objects (instances). As an example, “Laying the formwork boards” b is-instance-ofN “Construction”. • Cause-effect relations A number of causative verbs (also called lexical causatives) were used to describe the cause and effect relationships between concepts. These causative verbs include: b performs N, bproduces N, b harms N, b minimizes N, and b designs N. As an example, “Task bproduces N Ifc

Fig. 9. Problem description.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

10

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

Fig. 10. Context.

Building Element” can be paraphrased as “Task causes Ifc Building Element to be produced.” • Locative/spatial relations The relations were used to specify how some concept/object is located in space in relation to some reference concept/object. For example, Worker b works_onN Work Platform.

4.4. Attributes Concepts alone do not provide enough information to achieve the purpose of the AFPS-Onto. Designing an AFPS is an information-intensive task and much of the required information is located as attributes of concepts. According to El-Gohary and El-Diraby [25], an “attribute” is defined as “a characteristic that describes a thing”. For example, the information about “worst-case factored effect of the maximum load” and “minimum factored resistance” is required when a PE assesses whether a component (e.g., sling) meets relevant requirements. Fig. 6 lists key attributes of main concepts. Values of attributes take different data types, including Integer, Float, String, and Boolean. These attributes in the ontology are grouped as slots attached to specific concepts. One of the benefits is that PEs can better determine the values of attributes and utilize them for different purposes, such as case retrieval process and automated design. It should be noted that some attributes of classes were not directly taken from design standards, but derived from the knowledge engineering process. For example, five attributes were tested to be critical to

determine which type of active fall system should be used for a given working-at-height problem, namely “movement direction”, “anchorage shape”, “work platform shape” “relative position between anchorage and work platform”, and “leading edge”. Values of “relative position between anchorage and work platform” include: (1) anchorage above platform, (2) anchorage adjacent to platform, (3) anchorage below platform, (4) anchorage on platform, and (5) anchorage separate from platform. Different values were determined according to the shape of work platform. For the work platform in the shape of plane (e.g., floor), possible values that the attribute “leading edge” takes on are: (1) fully open, (2) three-side open, (3) adjacent half open, (4) opposite half open, (5) one-side open, and (6) fully closed. For the work platform in the shape of line (e.g., I-beam), the attribute has the following values: (1) full open and (2) half open. These attributes play important roles in facilitating reasoning process in knowledge-based systems.

4.5. Axioms Axioms are essential to define the semantics of concepts and relations [25]. They provide a declarative specification for the definitions of and constraints on the concepts and relations. To serve its purposes, the AFPS-Onto must contain a necessary and sufficient set of axioms for representation and reasoning. The specifications are represented in First Order Logic. Axioms allow a knowledge-based system to interpret the information. The following are some examples of axioms used in the AFPS-Onto.

Fig. 11. System layout.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

11

Fig. 12. Calculation of single span (before a fall).

Fig. 13. Calculation of single span (after a fall).

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

12

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

Table 5 Components and minimum requirements. Item

Quantity Applicable design standards

Anchorage subsystem [Steel pipe] 2

Minimum factored resistance

Worst-case factored effect of the maximum load

Size/length

Others

N.A.

Shear force:20.8 kN Moment: 31.2 kNm

Shear force: 15.60 kN Moment:23.40 kNm

Shear force: 53.55 kN Moment: / Shear force: 23.56 kN Moment: / Shear force: 23.56 kN Moment: / Shear force: / Moment: 12 kNm

Shear force: 17.67 kN Moment: / Shear force: 17.67 kN Moment: / Shear force: 17.67 kN Moment: / Shear force: / Moment: 9.0 kNm

Size: 100 mm ∗ 100 mm; Thickness: 9 mm; Length: 2.5 m Diameter:16 mm Length: at least 8 m

E: 400 MPa Weight: 1.55 N/m

Shear force: / Moment: 12.0 kNm Shear force: / Moment: 18.0 kNm

Shear force: / Moment: 9.0 kNm Shear force: / Moment: 9.0 kNm

Horizontal lifeline Sling

1

[SS 570]

2

[EN 795]

Connector

2

[EN 362]

Bull ring

1

N.A.

Personal fall arrest system Hook 1

[SS 528-5]

Lanyard

1

[SS 528-2]

PEA

1

[SS 528-2]

Shear force: / Moment: 12.0 kNm

Shear force: / Moment: 9.0 kNm

Karabiner

1

[SS 528-5]

Full body harness

1

[SS 528-1]

Shear force: / Moment: 12.0 kNm N.A.

Shear force: / Moment: 9.0 kNm N.A

4.5.1. Active fall protection system • A vertical lifeline system is an active fall protection system that has subsystems of a vertical lifeline, a fall arrester, connecting devices, and body holding device, (∀ x, s1, s2, s3, s4) ((active_fall_protection_system (x) ∧ has_subsystem (x, s1, s2, s3, s4) ∧ vertical_lifeline (s1) ∧ fall_arrester (s2) ∧ connecting_devices (s3) ∧ body_holding_device (s4) ⊃ vertical_lifeline_system (x)). • A travel restraint system is a system that prevents one or more workers from reaching an unprotected edge or opening. It couples the workers' body-holding devices to an anchorage using a suitable means, such as restraint lanyards. (∀ x, s1, s2, s4) ((active_fall_protection_system (x) ∧ has_subsystem (x, s1, s2, s3) ∧ restraint_anchorage (s1) ∧ restraint_lanyard (s2) ∧ body_holding_device (s3) ∧ restricts_movement_of (x, w) ∧ worker (w) ⊃ travel_restraint_system (x)). 4.5.2. Fall protection equipment • A restraint anchorage is an anchorage that is a subsystem of a travel restraint system, (∀ x, y) ((anchorage (x) ∧ travel_restraint_system (y) ∧ is_subsystem_of (x, y)) ⊃ restraint_anchorage (x)).

Diameter:12.7 mm Length: 0.8 m (with PEA)

E: 450 MPa

Max force: 6 kN Average force: 3.2 kN Max deployment: 1.4 m

Appropriate size for users

4.6. Coding The AFPS-Onto was coded using Protégé 5.0 (see Fig. 7). Protégé is a free, open-source ontology editor and framework for building intelligent systems [70]. Concepts, relations, and attributes were modelled as “classes”, “object properties” and “data properties”, respectively. Axioms were represented in Protégé through the use of OWL restrictions (i.e., quantifier restrictions, cardinality restrictions, and has_value restrictions), characteristics of object property, and datatype restrictions. 5. Ontology evaluation Ontology evaluation is an essential for the development of ontologies. It represents a “judgement of the ontology content with respect to a particular frame of reference” [25]. A number of ontology evaluation approaches exist, such as gold standard evaluation, data driven evaluation, automated consistency checking, criteria-based evaluation, evaluation by humans, application-based evaluation and task-based evaluation [53,71–73]. However, it is important to adopt appropriate formal evaluation criteria and approaches, since some may not fit well to the

Fig. 14. Residual risk.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

ontology and its application domain. A summary of ontology evaluation approaches is presented in Table 4. Note that when selecting evaluation approaches, one should consider the limitations of each approach and make sure that adopted evaluation approaches match the objectives of the AFPS-Onto. In this study, the gold standard evaluation approach was not selected because there is no existing benchmark ontology in the AFPS domain. Although the data driven approach is powerful to evaluate the coverage of ontologies, it is not suitable to assess the correctness, clarity, and applicability of an ontology [53]. As a result, this study adopted a combination of automated consistency checking, criteria-based evaluation, and task-based evaluation for evaluating the AFPS-Onto. 5.1. Automated consistency checking By definition, consistency checking ensures that an ontology does not include any contradictory facts. For definitions to be inferentially and semantically consistent, they must be able to obtain consistent conclusions using the meaning of all definitions and axioms [74]. The AFPSOnto was first evaluated by using the Description Logic reasoner: Pellet. Pellet can not only detect unsatisfactory concepts (or inconsistent ontologies) but also support the diagnosis and resolution of the bug. Pellet has competitive advantages in terms of the time it takes the reasoner to check the consistency of the data, the amount of total preparation time spent before queries can be answered, and the time it takes to answer each query [75]. As a third-party plug-in, Pallet was invoked by choosing Pellet from Reasoner options. The Protégé then automatically checked the consistency of the ontology. The inferred classes and relations were then also shown on the interface. The reasoner checked for hierarchies, domains, ranges, and conflicting disjoint assertions. Inconsistent classes are marked with red. The automated consistency checking process was iterative, as the ontology was developed incrementally by adding new definitions and modifying old ones. 5.2. Criteria-based evaluation Criteria-based evaluation is mainly focused on verifying the content of an ontology. Four core criteria which match the objectives of the AFPS-Onto were selected from evaluation criteria proposed by past studies [20,55,56], including clarity, extendibility, completeness, and coverage. • Clarity — Clarity refers to whether an ontology effectively communicates the intended meaning of defined terms, whether definitions in the ontology are clearly specified without ambiguity [55]. This criterion was considered as early as the conceptualization phase. To ensure the clarity of the ontology, most concepts and their definitions were extracted directly from the design standard “Specification for design of active fall-protection systems, SS 607” by Singapore Standards Council [9]. The Standard was prepared by a working group appointed by the Technical Committee on Personal Safety and Health under the direction of the Singapore Standard Council. The ontology concepts were deemed to be defined formally and unambiguously. For example, “Anchorage subsystem” was defined as “A subsystem of a complete active fall-protection system to which workers connect their personal equipment.” In addition, different concepts used in different Standards are synonyms. For example, “Professional engineer” used in SS 607 [9] is a synonym of the “Qualified engineer” used in Z259.1615. To clarify, synonyms of a given concept were listed in the Protégé through the use of brdfs:comment N, an instance of b rdf:property N that may be used to provide a human-readable description of a resource. • Extendibility — Extendibility, also called expandability, refers to an ontology's ability to expand itself to “describe specific application domains in a way that does not change the current definitions within the ontology” [55]. As discussed earlier in this paper, the AFPS-Onto was

13

built to be scalable so that it can describe other application domains such as passive fall protection system design (e.g., guardrail systems and nets) and control measures of other hazards (e.g., struck by moving objects, caught-in/between, and electrocutions). The extension can be made based on the generic ontological model (see Fig. 3), without changing the current definitions of concepts in the AFPS-Onto. For example, if the ontology is extended by including passive fall protection system design, such an extension would not require to change well-defined concepts and relations in the AFPS-Onto. • Completeness — This criterion refers to whether definitions of concepts are complete. Incompleteness is a serious problem in ontologies [74]. In fact, it is not possible to prove both the completeness and incompleteness of an ontology. However, according to Gómez-Pérez [74], an ontology is considered semantically complete if it meets the two requirements: (1) each definition is complete and (2) the ontology explicitly includes all that is supposed to be included. The incompleteness can be determined when a concept is not explicitly defined and inferred using other axioms and definitions. The incompleteness of the ontology can also be detected by considering its reference to the real world itself, as suggested by Yu et al. [55]. Based on GómezPérez's suggestions [74], the following three steps were taken to identify incompleteness of definition. The first step was to check completeness of the class hierarchy in which the current definition is included. Imprecise and over-specified classes were modified. For example, a restraint anchorage is not only a subclass of the “Anchorage” but also is a subsystem of the “Travel restraint system”. The ontology would not be complete if the latter definition was missing. The second step was to check the completeness of the domains and ranges of the functions and relations and that the domains of these functions and relations are defined in the class hierarchy of the AFPS-Onto. The last step was concerned with checking the completeness of the classes. • Coverage — This criterion refers to the coverage of the concepts over the domain of active fall-protection system design. In order to ensure the coverage of the AFPS-Onto during the development process, core concepts and attributes used in relevant design Standards were manually extracted and organized. The coverage of the ontology was evaluated by checking whether these core concepts and attributes were included in the ontology. It was undertaken by using the information collected from interviews and design standards as a frame of reference to identify incompleteness of the ontology in terms of scope, exhaustiveness, and granularity. The coverage was improved by adding missing core concepts and attributes to the ontology.

5.3. Task-based evaluation To evaluate the competency of the AFPS-Onto, it (relevant concepts, relations, and attributes) was used to describe the content of design cases, including problem description, context, solution, test, and residual risk. The ontology was evaluated by assessing its competency to describe forty-one complete and real AFPS design cases. For illustration purposes, only one from the forty-one design cases was illustrated below using the AFPS-Onto. Fig. 8 describes a hazardous working-at-heights scenario in Singapore, in which a number of workers were laying steel pipes on the formwork for concrete casting. Note that the picture alone does not provide sufficient information for designing an appropriate and effective AFPS to protect workers. A fall arrest system was designed and the design report was created based on the AFPS-Onto. The design report consists of: (1) problem description (see Fig. 9), (2) context (see Fig. 10), (3) system layout (see Fig. 11), (4) calculations (see Figs. 12 and 13, and Table 5), and (5) residual risk (see Fig. 14). With reference to Fig. 9, classes are highlighted in bold with uppercase of the first letter (e.g., Worker), individuals of a class are

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

14

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

highlighted with a pair square brackets (e.g., [laying steel pipes]), relations are highlighted with a pair of angle brackets (e.g., bperformsN), attributes are highlighted in Italic (e.g., movement direction), and values of attributes are highlighted with underline in Italic (e.g., adjacent to). 5.4. Evaluation results Automated consistency checking has been objective and straightforward. Inconsistencies were corrected along the development process of the AFPS-Onto based on the feedback from Pallet. Final results of the automated consistency checking were positive and no inconsistencies were detected. Results of criteria-based evaluation indicated that the AFPS-Onto was clear, extendable, and complete in terms of scope, exhaustiveness, and granularity. Noted that clarity and completeness are closely connected. An ontology can be complete and not be clear when the ontology includes redundancies which may hinder the inference from its definitions and axioms. On the other hand, an ontology can be clear but not complete when it cannot be inferred because definitions of classes and relations are incomplete. In addition, coverage of concepts in an ontology should be assessed based on its purposes. The AFPS-Onto captures all essential concepts which serve the purpose of AFPS design. But it is incomplete when it is used for fall protection safety management on site. The coverage and usefulness were further proved by the task-based evaluation. The ontology was used to describe forty-one AFPS design cases. Results suggested that the ontology is capable of capturing key concepts and relations within the cases. This suggests that the AFPSOnto can be used to standardize the structure of AFPS design report and facilitate knowledge reuse and storage.

6. Conclusions Construction sites are hazardous in nature and it is important to systematically identify and manage core hazards. Designing hazard control measures is an information intensive task, which would benefit from using knowledge-based systems. This paper represents an initial effort to develop a global ontology of construction safety hazard management (CSHM-Onto). It adopted a bottom-up strategy by developing and evaluating an ontology (AFPS-Onto) for the design of active fall protection system in the building and construction industry. Following the definition provided by [20,18], the AFPS-Onto can be defined as a formal, explicit specification of a shared conceptualization of AFPS design domain. A ‘conceptualization’ refers to an abstract model of AFPS design based on relevant concepts, relations, and axioms within the domain. ‘Explicit’ means the type of concepts is explicitly defined. ‘Formal’ means that the AFPS-Onto is machine readable. ‘Shared’ represents the notion that the AFPS-Onto reflects a certain rate of consensual knowledge accepted by people within the domain. It is an engineering artefact, constituted by a hierarchy of concepts, attributes, and relations among concepts used to represent the knowledge of AFPS design. The ontology builds itself on a generic ontological model which can be used as a harmonization framework within which a hierarchy of hazard control measure ontologies can be built and combined. The framework consists of three main parts (i.e., problem, context, and solution) and nine concepts (i.e., task, actor, ifc building element, hazard, construction method, constraint, safety resource, hazard control measure, and residual risk). The ontology was evaluated based on automated consistency checking, criteria-based evaluation, and task-based evaluation. Results indicated that the AFPS-Onto is consistent, credible and effective in capturing and describing relevant knowledge required in design of AFPS. The core contribution of the ontology is two-fold. First, the ontology fills the knowledge gap by providing a formal and shared vocabulary for the domain of AFPS design. By using the ontology, PEs can standardize design reports and thus avoid unnecessarily misunderstanding caused

by inconsistent terms and assumptions. In doing so, AFPS knowledge reuse and sharing can be promoted. Second, the ontology can be used to develop decision support systems of AFPS design. In specific, it can be utilized to define query vocabulary, build case structure, and design rules. The research team is developing a knowledge-based system for supporting the design of AFPS based on the ontology. In addition, by integrating the concepts of IFC building element, it is possible to integrate the system into building information modelling (BIM) for visualization and planning. 7. Limitations and future work The ontology developed in this paper has the following limitations. Firstly, it focuses only on knowledge representation of the domain of active fall protection system design. As aforementioned, PPE is ranked as the least effective method to managing hazards. In most working-atheights scenarios, elimination and engineering controls should first be considered. Nevertheless, many work at heights scenarios still require PPE as a last line of defense. Secondly, the ontology represents an effort to construct a model of AFPS design domain in part based on knowledge-engineering process. As a domain ontology, the AFPS-Onto is neither the only ontology nor perfect one for the domain of AFPS design. In fact, a domain of interest cannot be fully represented by a single ontology [20,25]. In addition, ‘perfect’ ontologies do not exist for a certain domain [76]. As such, the reliability and validity of the study should be assessed based on the nature and purpose of ontology. Reliability and replication are not pertinent to ontology engineering, since it is neither necessary nor possible to replicate an ontology using the same methodology and data. Nevertheless, this study established confidence in the soundness and usefulness of the AFPS-Onto with respect to its purpose. As the ontology was largely based on AFPS design standards from the US, Canada, and Singapore, results of this study can be generalized across these areas. The last limitation is associated with ontology evaluation methods. While the AFPS-Onto was evaluated objectively based on automated consistency checking and task-based evaluation, criteria-based evaluation was largely a subjective approach, as criteria like clarity, extendibility, and completeness are hard to quantify. However, coverage can be objectively evaluated by comparing the ontology with a corpus of information (e.g., relevant design standards and real design case reports). Measures can be used to quantify coverage, including number of overlapping terms, vector space similarity, precision, recall, and Fmeasure [55,72]. The ontology places itself on the local level. Future effort can be made to develop other local ontologies of other control measures against fall from heights (e.g., passive fall protection systems). These local ontologies could be integrated and mapped into a fall from heights (FFH) ontology at the intermediate-level. The basic idea is to develop a library of construction hazard management ontologies in a standard formalism and design and share knowledge bases across people and computer systems. Ontology-based expert systems (e.g., rule-based or Bayesian-network-based) can be developed to help safety practitioners manage hazards in a smarter way. However, it should be noted that the evolvability of the ontologies at local and intermediate levels does not guarantee the evolvability of expert systems which use the ontologies. This is because knowledge bases developed at the local level (e.g., the AFPS design domain) may not be valid at the intermediate level (e.g., fall from heights). Acknowledgments This work was granted by the Workplace Safety and Health Institute, Singapore, and the Ministry of Manpower, Singapore (grant number—MOMOSDETT12000007). The authors cordially thank the funding agency.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

Reference [1] The British Standards Institution, Code of Practice for the Design of Buildings Incorporating Safe Work at Height, BSI Standards Limited, London, 2012. [2] United States Department of Labor, National Census of Fatal Occupational Injuries in 2014, 2015 (available via http://www.bls.gov). [3] Health and Safety Executive, Health and Safety in Construction in Great Britain, 2014, London, 2014 (available via http://www.hse.gov.uk/statistics/industry/ construction). [4] Safe Work Australia, in: S.W. Australia (Ed.), Work-related injuries and fatalities involving a fall from height, Australia, Safe Work Australia, Canberra, Australia, 2013. [5] Statistics New Zealand, Injury Statistics, in: S.N. Zealand (Ed.), Work-related Claims: 2013 – Trends Tables for 2002–12, 2013 (Wellington, New Zealand). [6] Workplace Safety and Health Institute, Workplace Safety and Health Report 2015, 2016 (available via https://www.wsh-institute.sg). [7] H. Hsiao, P. Simeonov, Preventing falls from roofs: a critical review, Ergonomics 44 (5) (2001) 537–561, http://dx.doi.org/10.1080/00140130110034480. [8] J. Sa, D.-C. Seo, S.D. Choi, Comparison of risk factors for falls from height between commercial and residential roofers, J. Saf. Res. 40 (1) (2009) 1–6, http://dx.doi. org/10.1016/j.jsr.2008.10.010. [9] Singapore Standards Council, Specification for Design of Active Fall-protection Systems SS 607 2015, Spring Singapore, Singapore, 2015. [10] Y.M. Goh, P.E. Love, Adequacy of personal fall arrest energy absorbers in relation to heavy workers, Saf. Sci. 48 (6) (2010) 747–754, http://dx.doi.org/10.1016/j.ssci. 2010.02.020. [11] Canadian Standards Association, Z259.16-15 Design of Active Fall-protection Systems, Canadian Standards Association, Mississauga, Ontario, Canada, 2015. [12] American National Standards Institute, American Society for Safety Engineers, ANSI/ ASSE Z359.6-2009 Specifications and Design Requirements for Active Fall Protection Systems, ANSI/ASSE, New York, 2009. [13] Y.M. Goh, Q. Wang, Investigating the adequacy of horizontal lifeline system design through case studies from Singapore, J. Constr. Eng. Manag. 141 (7) (2015) 04015017, http://dx.doi.org/10.1061/(ASCE)CO.19437862.0000989. [14] Y.P. Hoe, Y.M. Goh, S.Y. Sim, Design of Fall Arrest Systems: A Review of the Current Issues in the Singapore Construction Industry, CIB W99 International Conference, Modelling and Building Health and Safety, Singapore, 2012. [15] S. Kumar, R. Singh, An automated design system for progressive die, Expert Syst. Appl. 38 (4) (2011) 4482–4489, http://dx.doi.org/10.1016/j.eswa.2010.09.121. [16] S.K. Chandrasegaran, K. Ramani, R.D. Sriram, I. Horváth, A. Bernard, R.F. Harik, W. Gao, The evolution, challenges, and future of knowledge representation in product design systems, Comput. Aided Des. 45 (2) (2013) 204–228, http://dx.doi.org/10. 1016/j.cad.2012.08.006. [17] S. Trausan-Matu, Ontology-based Interoperability in Knowledge-based Communication Systems, Ontologies in Urban Development Projects, Springer, 2011 139–152. [18] R. Studer, V.R. Benjamins, D. Fensel, Knowledge engineering: principles and methods, Data Knowl. Eng. 25 (1) (1998) 161–197, http://dx.doi.org/10.1016/ S0169-023X(97)00056-6. [19] O. Corcho, M. Fernández-López, A. Gómez-Pérez, Methodologies, tools and languages for building ontologies. Where is their meeting point? Data Knowl. Eng. 46 (1) (2003) 41–64, http://dx.doi.org/10.1016/S0169-023X(02)00195-7. [20] T.R. Gruber, Toward principles for the design of ontologies used for knowledge sharing? Int. J. Hum. Comput. Stud. 43 (5) (1995) 907–928, http://dx.doi.org/10.1006/ ijhc.1995.1081. [21] M. Uschold, M. Gruninger, Ontologies: principles, methods and applications, Knowl. Eng. Rev. 11 (2) (1996) 93–136, http://dx.doi.org/10.1017/S0269888900007797. [22] R. Neches, R.E. Fikes, T. Finin, T. Gruber, R. Patil, T. Senator, W.R. Swartout, Enabling technology for knowledge sharing, AI Mag. 12 (3) (1991) 36. [23] B. Chandrasekaran, J.R. Josephson, V.R. Benjamins, What are ontologies, and why do we need them? IEEE Intell. Syst. 14 (1) (1999) 20–26, http://dx.doi.org/10.1109/ 5254.747902. [24] S. Grimm, P. Hitzler, A. Abecker, Knowledge representation and ontologies: logic, ontologies and semantic web language, Semantic Web Services, Springer, 2007. [25] N.M. El-Gohary, T.E. El-Diraby, Domain ontology for processes in infrastructure and construction, J. Constr. Eng. Manag. 136 (7) (2010) 730–744, http://dx.doi.org/10. 1061/(ASCE)CO.1943-7862.0000178. [26] D.B. Lenat, R.V. Guha, Building Large Knowledge-based Systems: Representation and Inference in the Cyc Project, Addison-Wesley, Boston, 1990. [27] The British Standards Institution, Geographic Information — Ontology — Part 1: Framework, BSI Standards Limited, London, 2015. [28] The British Standards Institution, Geographic Information — Ontology — Part 2: Rules for Developing Ontologies in the Web Ontology Language (OWL), BSI Standards Limited, London, 2015. [29] Z. Zhou, Y.M. Goh, L. Shen, Overview and analysis of ontology studies supporting development of the construction industry, J. Comput. Civ. Eng. (2016) 04016026, http://dx.doi.org/10.1061/(ASCE)CP.1943-5487.0000594. [30] M. Grüninger, M.S. Fox, Methodology for the design and evaluation of ontologies, IJCAI95, Workshop on Basic Ontological Issues in Knowledge Sharing, Montreal, 1995. [31] M. Uschold, M. King, Towards a methodology for building ontologies, IJCAI95 Workshop on Basic Ontological Issues in Knowledge Sharing, Montreal, 1995. [32] M. Fernández-López, A. Gómez-Pérez, N. Juristo, Methontology: from ontological art towards ontological engineering, Proc., Ontological Engineering Spring Symp. Series, American Association for Artificial Intelligence, Menlo Park, CA, 1997. [33] N.F. Noy, D.L. McGuinness, Ontology Development 101: A Guide to Creating your First Ontology, Stanford Knowledge Systems, Stanford, CA, 2001. [34] S. Staab, R. Studer, H.-P. Schnurr, Y. Sure, Knowledge processes and ontologies, IEEE Intell. Syst. 16 (1) (2001) 26–34, http://dx.doi.org/10.1109/5254.912382.

15

[35] S. Fenz, An ontology-based approach for constructing Bayesian networks, Data Knowl. Eng. 73 (2012) 73–88, http://dx.doi.org/10.1016/j.datak.2011.12.001. [36] A. Devitt, B. Danev, K. Matusikova, Constructing Bayesian networks automatically using ontologies, Proceedings of Second Workshop on Formal Ontologies Meets Industry (FOMI 2006), 2006. [37] A. Grubišić, S. Stankov, I. Peraić, Ontology based approach to Bayesian student model design, Expert Syst. Appl. 40 (13) (2013) 5363–5371, http://dx.doi.org/10.1016/j. eswa.2013.03.041. [38] A. Yurchyshyna, A. Zarli, An ontology-based approach for formalisation and semantic organisation of conformance requirements in construction, Autom. Constr. 18 (8) (2009) 1084–1098, http://dx.doi.org/10.1016/j.autcon.2009.07.008. [39] M. Dibley, H. Li, Y. Rezgui, J. Miles, An ontology framework for intelligent sensorbased building monitoring, Autom. Constr. 28 (2012) 1–14, http://dx.doi.org/10. 1016/j.autcon.2012.05.018. [40] P. Zhou, N. El-Gohary, Ontology-based automated information extraction from building energy conservation codes, Autom. Constr. 74 (2017) 103–117, http://dx. doi.org/10.1016/j.autcon.2016.09.004. [41] D.-Y. Lee, H.-l. Chi, J. Wang, X. Wang, C.-S. Park, A linked data system framework for sharing construction defect information using ontologies and BIM environments, Autom. Constr. 68 (2016) 102–113, http://dx.doi.org/10.1016/j.autcon.2016.05.003. [42] S.-K. Lee, K.-R. Kim, J.-H. Yu, BIM and ontology-based approach for building cost estimation, Autom. Constr. 41 (2014) 96–105, http://dx.doi.org/10.1016/j.autcon. 2013.10.020. [43] B. Zhong, L. Ding, H. Luo, Y. Zhou, Y. Hu, H. Hu, Ontology-based semantic modeling of regulation constraint for automated construction quality compliance checking, Autom. Constr. 28 (2012) 58–70, http://dx.doi.org/10.1016/j.autcon.2016.08.027. [44] S. Kaza, H. Chen, Public safety information sharing: an ontological perspective, Digital Government, Springer 2008, pp. 263–282. [45] Q.T. Le, D.Y. Lee, C.S. Park, A social network system for sharing construction safety and health knowledge, Autom. Constr. 46 (2014) 30–37, http://dx.doi.org/10. 1016/j.autcon.2014.01.001. [46] H.-H. Wang, F. Boukamp, Ontology-based representation and reasoning framework for supporting job hazard analysis, J. Comput. Civ. Eng. 25 (6) (2011) 442–456, http://dx.doi.org/10.1061/(ASCE)CP.1943-5487.0000125. [47] N.-W. Chi, K.-Y. Lin, S.-H. Hsieh, Using ontology-based text classification to assist job hazard analysis, Adv. Eng. Inform. 28 (4) (2014) 381–394, http://dx.doi.org/10. 1016/j.aei.2014.05.001. [48] S. Zhang, F. Boukamp, J. Teizer, Ontology-based semantic modeling of construction safety knowledge: towards automated safety planning for job hazard analysis (JHA), Autom. Constr. 52 (2015) 29–41, http://dx.doi.org/10.1016/j.autcon.2015. 02.005. [49] Y. Lu, Q. Li, Z. Zhou, Y. Deng, Ontology-based knowledge modeling for automated construction safety checking, Saf. Sci. 79 (2015) 11–18, http://dx.doi.org/10.1016/ j.ssci.2015.05.008. [50] K. Hu, H. Rahmandad, T. Smith-Jackson, W. Winchester, Factors influencing the risk of falls in the construction industry: a review of the evidence, Constr. Manag. Econ. 29 (4) (2011) 397–416, http://dx.doi.org/10.1080/01446193.2011.558104. [51] Y.M. Goh, N.F. Binte Sa'adon, Cognitive factors influencing safety behavior at height: a multimethod exploratory study, J. Constr. Eng. Manag. 141 (6) (2015) 04015003, http://dx.doi.org/10.1061/(ASCE)CO.1943-7862.0000972. [52] B.H.W. Guo, T.W. Yiu, V.A. González, Predicting safety behavior in the construction industry: development and test of an integrative model, Saf. Sci. 84 (2016) 1–11, http://dx.doi.org/10.1016/j.ssci.2015.11.020. [53] P.D. Haghighi, F. Burstein, A. Zaslavsky, P. Arbon, Development and evaluation of ontology for intelligent decision support in medical emergency management for mass gatherings, Decis. Support. Syst. 54 (2) (2013) 1192–1204, http://dx.doi.org/10. 1016/j.dss.2012.11.013. [54] C. Roussey, F. Pinet, M.A. Kang, O. Corcho, An introduction to ontologies and ontology engineering, Ontologies in Urban Development Projects, Springer 2011, pp. 9–38. [55] J. Yu, J.A. Thom, A. Tam, Evaluating ontology criteria for requirements in a geographic travel domain, OTM Confederated International Conferences on the Move to Meaningful Internet Systems, Springer 2005, pp. 1517–1534. [56] A. Gómez-Pérez, N. Juristo, J. Pazos, Evaluation and assessment of knowledge sharing technology, in: N.J.I. Mars (Ed.), Towards very Large Knowledge Bases, IOS Press, Amsterdam, Netherlands 1995, pp. 289–296. [57] BuildingSMART International Limited, Industry Foundation Classes Version 4 - Addendum 1(available via http://www.buildingsmart-tech.org/ifc/IFC4/Add1/html/ toc.htm) 2014. [58] M. Arashpour, R. Wakefield, N. Blismas, J. Minas, Optimization of process integration and multi-skilled resource utilization in off-site construction, Autom. Constr. 50 (2015) 72–80, http://dx.doi.org/10.1016/j.autcon.2014.12.002. [59] M. Arashpour, R. Wakefield, B. Abbasi, E. Lee, J. Minas, Off-site construction optimization: sequencing multiple job classes with time constraints, Autom. Constr. 71 (2016) 262–270, http://dx.doi.org/10.1016/j.autcon.2016.08.001. [60] L. Jaillon, C. Poon, Life cycle design and prefabrication in buildings: a review and case studies in Hong Kong, Autom. Constr. 39 (2014) 195–202, http://dx.doi.org/10. 1016/j.autcon.2013.09.006. [61] N. Blismas, C. Pasquire, A. Gibb, Benefit evaluation for off-site production in construction, Constr. Manag. Econ. 24 (2) (2006) 121–130, http://dx.doi.org/10.1080/ 01446190500184444. [62] Building and Construction Authority Singapore, Prefabricating the Future - A Local CLT and GLULAM Experience(available via https://www.bcaa.edu.sg) 2016. [63] Building and Construction Authority Singapore, More Prefabricated Prefinished Volumetric Construction (PPVC) Projects Coming Up, Build Smart, Building and Construction Authority Singapore, Singapore, 2015.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009

16

B.H.W. Guo, Y.M. Goh / Automation in Construction xxx (2017) xxx–xxx

[64] Y.M. Goh, S. Chua, Knowledge, attitude and practices for design for safety: a study on civil & structural engineers, Accid. Anal. Prev. (2015)http://dx.doi.org/10.1016/j.aap. 2015.09.023. [65] Ministry of Manpower Singapore, Workplace Safety and Health Act(available via http://www.mom.gov.sg/workplace-safety-and-health/workplace-safety-andhealth-act) 2016. [66] Singapore Statutes Online, Workplace Safety and Health (Work at Heights) Regulations 2014(available via http://statutes.agc.gov.sg) 2014. [67] C.S. Khoo, J.-C. Na, Semantic relations in information science, Annu. Rev. inform. Sci. 40 (2006) 157, http://dx.doi.org/10.1002/aris.1440400112. [68] S. Purao, V.C. Storey, A multi-layered ontology for comparing relationship semantics in conceptual models of databases, Appl. Ontol. 1 (1) (2005) 117–139. [69] T. El-Diraby, H. Osman, A domain ontology for construction concepts in urban infrastructure products, Autom. Constr. 20 (8) (2011) 1120–1132, http://dx.doi.org/10. 1016/j.autcon.2011.04.014. [70] M. Horridge, H. Knublauch, A. Rector, R. Stevens, C. Wroe, A Practical Guide to Building OWL Ontologies Using the Protégé-OWL Plugin and CO-ODE Tools Edition 1.0, University of Manchester, 2004.

[71] J. Yu, J.A. Thom, A. Tam, Ontology evaluation using Wikipedia categories for browsing, Proceedings of the Sixteenth ACM Conference on Conference on Information And Knowledge Management, ACM 2007, pp. 223–232. [72] C. Brewster, H. Alani, S. Dasmahapatra, Y. Wilks, Data driven ontology evaluation, Proc. of the 4th International Conference on Language Resources and Evaluation, Lisbon, European Language Resources Association, 2004. [73] J. Brank, M. Grobelnik, D. Mladenic, A survey of ontology evaluation techniques, Proceedings of the Conference on Data Mining and Data Warehouses (SiKDD 2005) 2005, pp. 166–170. [74] A. Gómez-Pérez, Towards a framework to verify knowledge sharing technology, Expert Syst. Appl. 11 (4) (1996) 519–529, http://dx.doi.org/10.1016/S09574174(96)00067-X. [75] E. Sirin, B. Parsia, B.C. Grau, A. Kalyanpur, Y. Katz, Pellet: a practical owl-dl reasoner, Web Semant. Sci. Serv. Agents World Wide Web 5 (2) (2007) 51–53. [76] A. Taivalsaari, Classes vs. prototypes: Some philosophical and historical observations, J. Object Orient Prog. 10 (7) (1996) 44–50.

Please cite this article as: B.H.W. Guo, Y.M. Goh, Ontology for design of active fall protection systems, Automation in Construction (2017), http:// dx.doi.org/10.1016/j.autcon.2017.02.009