1 Developing a Conceptual Framework of Smart Work Packaging for Constraints 2 Management in Prefabrication Housing Production 3 This is the peer-reviewed post-print version of the paper: Li, X., Shen, G. Q., Wu, P., Xue, F., Chi, H. L., & Li, C. Z. (2019). Developing a conceptual framework of smart work packaging for constraints management in prefabrication housing production. Advanced Engineering Informatics, 42, 100938. Doi: 10.1016/j.aei.2019.100938 The final version of this paper is available at: https://doi.org/10.1016/j.aei.2019.100938. The use of this file must follow the Creative Commons Attribution Non-Commercial No Derivatives License, as required by Elsevier’s policy. 4 Highlights 5 6 7 8 • • • 9 Abstract 10 Constraints management is the process of satisfying bottlenecks to facilitate tasks assigned to 11 crews being successfully executed. However, managing constraints is inherently challenging in 12 prefabrication housing production (PHP), due to the fragmentation of processes and information 13 during project delivery. Enlightened by the broadly accepted work packaging method and the 14 smart construction objects (SCOs) model, this study aims to define and implement smart work 15 packaging (SWP) for constraints management in PHP. Firstly, the framework of SWP-enabled 16 constraints management (SWP-CM) with three primary functions, including constraints modeling, 17 constraints optimization, and constraints monitoring, is established. In addition, this study 18 develops a layered abstract model as a prototype representation to elaborate on the implementation 19 of SWP for practitioners. Finally, a laboratory-based test is applied to validate the framework. It 20 can prove that SWP indeed opens new avenues for smart constraints management for PHP. • Smart work packaging in prefabrication housing production is investigated. A framework using smart work packaging is established for constraint management. A layered system is proposed to achieve modeling, optimization, and monitoring. The validity of the proposed framework is demonstrated by a lab-based simulation game. 21 Keywords: Smart Work Packaging (SWP), Prefabrication Housing Production (PHP), Constraints 22 Management, Building Information Modeling (BIM) 23 1. Introduction 24 Prefabrication housing production (PHP) is an innovative approach that the prefabricated material, 25 components, modules, and units are manufactured efficiently at different locations and then 26 converge at the site for installation. This approach could alleviate the labor shortage and swiftly 27 provide housings to mitigate the unbalanced housing supply and demand in Hong Kong (Li et al., 28 2018a; Wu et al., 2017). Although PHP has proven to be useful in the supply of public rental 29 housing (PRH), it is still plagued by the pathological schedule delay which can lead to an adverse 30 impact on Hong Kong’s economic growth and competitiveness, particularly when manufacturing 31 plants have been moved to the Great Bay Area of Mainland China. For example, the government 32 planned to construct 13300 flat units of PRH in the financial year of 2016-2017. However, the 33 actual amount of PRH production is 11276 units, and 15.22% delay occurred (Housing Authority, 34 2018). Dominant drivers for such delays have proven to be the uncertainties and constraints (Li et 35 al., 2017a). Uncertainty means something that may occur, whereas constraint (e.g., limited space 36 and buffers) is something that will happen. The constraints are the obvious bottlenecks and are 37 more predictable than the uncertainties to be improved in task executions. Hence, reliable 38 constraint-free workflows are vital for achieving an industrialized PHP environment across design, 39 manufacturing, logistics, and on-site assembly so as to avoid schedule delays and cost overruns 40 (Wang et al., 2016a). 41 The reliability of PHP schedules can be enhanced via proactive constraints management, which is 42 the process of identifying, optimizing, and monitoring of bottlenecks to ensure that work package- 43 level tasks assigned to crews can be timely and accurately executed. They can be related to 44 technical sequencing, temporal/spatial limitations, and safety/quality concerns. Examples of such 45 constraints include incomplete BIM models, drawings and specifications, unavailability of 46 workforce, materials, prefabricated products, equipment and tools, shortage of temporary 47 structures, limited workspace, lack of work permits, unidentified safety and hazard issues, 48 uncontrolled environmental conditions (e.g., severe weather), untimely and inaccurate 49 transportation, and uncompleted quality control. Managing constraints in PHP processes is to 50 prepare more (e.g., on detailed and dynamic planning with lean solutions) and act fast (e.g., on 51 decision-making) using available information and knowledge. As such, the primary objective of 52 constraints management is to continually improve the reliability of workflow by guaranteeing that 53 accurate information is always available at the right time in the right format to the right person. 54 Currently, there have been numerous studies focusing on how to support decision makers and 55 collaborative workers with precise and timely information for task execution (Zhong et al., 2017; 56 Li et al., 2018b;). For instance, an internet of things (IoT)-enabled Building Information Modeling 57 (BIM) platform is developed with the support of smart construction objects (SCOs) by equipping 58 objects with information and communication technologies such as radio frequency identification 59 (RFID), and by using augmented reality (AR), and other sensing and tracking technologies (Li et 60 al., 2018c; Niu et al., 2016). Although Wang et al. (2016a) have made efforts to develop a 61 framework by considering the adoption of information technologies for constraints management 62 in oil and gas industry, there is so far no widely accepted approach for constraints management in 63 PHP. 64 The development of smart work packaging (SWP) in recent years seems to be adequate to address 65 the challenge. In PHP, there are a few studies which investigate the smart transformation of a group 66 of tasks (e.g., the lowest level in the work breakdown structure) based on the building systems of 67 product breakdown structure (PBS) by embedding the capabilities of visualizing, tracking, sensing, 68 computing, networking, and reacting. The smart transformation centers upon autonomy, adaptivity, 69 and sociability, which can facilitate better tasks execution by crews. For instance, the PHP 70 machinery (e.g., cranes) can be augmented with the autonomy to transport or hoist the 71 prefabricated products independently and without direct intervention from the surroundings (Chi 72 et al., 2012). In addition, the PHP planning approaches can be enhanced with adaptivity to be 73 capable of reacting resiliently through dynamic re-planning when constraints are not removed 74 (Abuwarda and Hegazy, 2016). SWP can also be strengthened with sociability to interact in a peer- 75 to-peer manner with other work packages or resources to collectively model the constraints 76 (Taghaddos et al., 2012). 77 This study proposes and validates a new framework of SWP for constraint management in PHP 78 based on the established theories of work packaging and SCOs (Isaac et al. 2017; Niu et al. 2016). 79 Work packaging can break down the PHP processes into manageable pieces to facilitate execution 80 of activities or tasks. SWP intends to improve the constraints during task executions in an 81 autonomous, adaptive, and optimal manner. e.g., automatic identification and analysis of 82 constraints and their interrelationships (Hamdi, 2013; Isaac et al. 2017), real-time sensing and 83 tracking constraints status (Liu et al. 2015), and optimal constraints improvement planning in a 84 dynamic manner (Abuwarda and Hegazy, 2016). SCOs are construction resources augmented with 85 smart characteristics of awareness, communicativeness, and autonomy using emerging 86 information technologies. However, SCOs are usually defined on single construction objects, 87 without considering construction project operations such as work packaging. Thus, the 88 development of SWP, as the integration of work packaging and SCOs, seems necessary and 89 imperative to improve constraints management in PHP. To improve the shortcomings in current 90 practices of constraints management, this study aims to develop a conceptual framework of SWP- 91 enabled constraints management (SWP-CM) in PHP. The concrete objectives of this research are 92 well explained below: (1) to define the SWP; (2) to establish the framework of SWP-CM; (3) to 93 propose a functional structure of SWP as a layered system model; and (4) to validate the SWP- 94 CM by a simulation game. 95 2. Background 96 2.1 Constraints Management 97 Constraints management (CM) is one of the critical strategies for production control and planning. 98 The concept of constraint was firstly introduced in 1984 as the theory of constraints (TOC) which 99 is a management philosophy for identifying the most critical bottleneck that prevents achieving a 100 goal and then systematically improving the constraint until it is no longer the bottleneck (Goldratt 101 and Cox, 1984). It assumes that each intricate system may comprise multi-connected activities, 102 there is at least one activity that acts as a constraint in the fully connected system, and the entire 103 process throughput can only be maximized when the constraint is improved. A corresponding 104 deduction is that spending more time on optimizing non-constraints activities cannot generate 105 significant benefits, and only improvements to the constraint will reach the goal. Thus, TOC aims 106 to offer an accurate and continuous focus on improving the current constraint until it no longer 107 confines the goal, at which point the focus moves to the next constraint. Constraints management 108 systems have proven to be more effective when compared to the reorder-point systems and material 109 requirements planning systems in the aspects of capacity management, inventory management, 110 and process improvement in the manufacturing industry. It is also argued that constraints 111 management can outperform the Just-in-time system owing to the more targeted nature of 112 improvement efforts in constraints (Boyd and Gupta, 2004). However, there is still no sound 113 approach to improve constraint management for achieving efficient collaborative working and 114 decision-making at crew-level task executions. It is mainly due to the fragmented process and 115 information in PHP, which may prevent the workers from agile constraints identification (Gong 116 et al. 2019), adaptive constraints improvement (Abuwarda and Hegazy, 2016), and real-time 117 constraints monitoring (Liu et al. 2015). 118 2.2 Work Packaging Method 119 TOC, to some extent, has similar philosophies as the Lean Construction. TOC uses its laser-like 120 focus to improve the capacity, while Lean Construction uses the broad-spectrum tools to eliminate 121 waste. In real practice, as PHP projects do not have infinite resources, an optimization process is 122 needed to identify and improve the most critical constraints. In this instance, TOC can work as an 123 efficient mechanism in prioritizing improvements for constraints, while Lean Construction can 124 offer a rich toolbox of improvement techniques. Thus, the combination of TOC and Lean 125 Construction may generate synergy on constraints management. The significance of integrating 126 Lean Construction with constraints management to issue executable work plans has also been 127 widely recognized by the construction industry. For example, work packaging is a planned, 128 executable process to strategically decompose the PHP scope into distinct and manageable pieces 129 with proper sizing and criteria. Each work package should be assigned to an individual supervisory 130 unit that is able to handle all its constraints. Therefore, the tasks should be separated into smaller 131 pieces (e.g., 500-2000 man-hours of work) so as the benefits outweigh the additional 132 administrative burden (Isaac et al., 2017). Additionally, the most frequently used criteria in work 133 packaging design include the type of prefabricated product, the workface in which the 134 prefabricated product is located, the specific physical location of the prefabricated product, and 135 the workflows (Ibrahim et al., 2009). The dependencies between tasks/activities included in 136 various work packages should also be considered. Whereas the PHP can be broken down into a 137 group of building systems (e.g., structure, envelope, partitions, services, and equipment) with a 138 hierarchical product structure (e.g., material, component, module, unit) in the design, the work 139 packaging in PHP can be defined by considering both product breakdown structure (PBS) and 140 work breakdown structure. One of the practical examples is advanced work packaging (AWP), 141 which was developed through the collaboration between the construction owners association of 142 Alberta (COAA) and the Construction Industry Institute (Hamdi, 2013). AWP uses a hierarchy of 143 engineering work packages (EWPs), construction work packages (CWPs) and installation work 144 packages (IWPs) to allow engineering and procurement planning to be driven by construction 145 sequencing. It breaks down the project processes into CWPs aligned with WBS. CWPs, in turn, 146 contain one or more IWPs. However, the direct implementation of AWP in PHP may be limited. 147 It works well in handling the complex mega project (e.g., oil and gas project), but its organizational 148 structure with CWP, EWP, and IWP is hierarchical and not flattened enough for PHP to improve 149 the efficiency of decision making and collaborative working (Li et al., 2019). Moreover, there are 150 also several significant limitations in the current work packaging methods for efficiently managing 151 constraints in PHP. Firstly, the process for identification and analysis of constraints and their 152 interrelationships is sluggish because the constraints are only discussed in look-ahead meetings 153 rather than in real-time manner (Hamdi, 2013; Isaac et al. 2017). In addition, constraints status is 154 untraceable and non-transparent due to the lack of sensing and tracking technologies for 155 monitoring (Liu et al. 2015). Constraints improvement planning is usually static without the 156 dynamic replanning ability (Abuwarda and Hegazy, 2016). Enlightened by the smartness of smart 157 construction object (SCO) (Niu et al., 2016), a more collaborative, autonomous, and adaptive 158 approach for constraints management through constraints modeling, monitoring, and optimization 159 may be possible. 160 2.3 Development of the Smart Work Packaging Method 161 Previous studies have made efforts to improve the smartness in the process management of 162 prefabricated construction. For instance, Wang et al. (2016a) developed a framework for total 163 constraints management in the oil and gas industry. However, information technologies were only 164 conceptually discussed in their framework, and there was no validation (e.g., a prototype system) 165 to demonstrate the smartness of the framework in constraints management implementation. In 166 addition, Li et al. (2018a) investigated the stakeholder-associated risks to improve the reliability 167 of phase-level scheduling. However, this study did not investigate constraints in the task-level plan, 168 which are more predictable than the risks at the phase level. The on-site assembly service, 169 developed by Li et al. (2018b), provided one of the services in the IoT-enabled BIM platform, 170 which is a critical part to support smart work packaging (SWP). However, the platform cannot 171 further divide the on-site assembly service into collaborative and manageable processes, therefore 172 providing relevant work packages in each of the processes. Li et al. (2017a) developed a simulation 173 game to test the learning effect of adopting information technologies and lean principles in 174 prefabrication housing production process. Based on Li et al. (2017a), this study tries to enhance 175 the work packaging method and constraints management in this simulation game to validate the 176 proposed conceptual framework. 177 Much effort has also been made in using cutting-edge information technologies to make work 178 packages smart (Ibrahim et al., 2009; Abuwarda and Hegazy, 2016). For example, Isaac et al. 179 (2017) developed algorithms for BIM which can be integrated with design structure matrix and 180 domain mapping matrix to automatically label relationships between prefabricated products and 181 their following sequence in which the prefabricated products should be assembled. Table 1 182 demonstrates a summary of the studies related to the development of SWP. As shown in Table 1, 183 the development of SWP has focused on the various aspects of constraints management, including 184 modeling, monitoring, and optimization. Some studies, although not directly using the name 185 “smart work packaging” or SWP, address the interaction between humans, resources, and the 186 environment with smartness using emerging technologies such as IoTs, wireless sensor networks, 187 big data, cloud computing, or other enabling technology to facilitate task execution. 188 189 190 Table 1 Studies related to Smart Work Packaging in Manufacturing Industry Research Function Zhang et al. (2018) Monitoring Wan et al. (2018) Optimization Luo et al. (2018) Modeling Kim (2018) Optimization Blanco-Novoa et al.(2018) Modeling Longo et al. (2017) Optimization/Mo deling Wang et al. (2017) Monitoring/Mode ling Peruzzini and Pellicciari (2017) Optimization Lu et al. (2017) Monitoring Ren et al. (2017) Modeling Wang et al. (2016b) Optimization Wang et al. (2016c) Optimization/Mo deling/Monitoring Ivanov et al. (2016) Optimization Seiger et al. (2015) Optimization/Mo deling/Monitoring Giner et al. (2012) Modeling Interpretations IoT-enabled active sensing system to assist operators in monitoring the real-time manufacturing process Cyber-physical production system (CPPS)enabled dynamic resource allocation for operators Using mobile intelligence to handle low-priority data to improve the data delivery efficiency Predefined jobs can be processed concurrently in different machines AR-based interface to assign tasks to the operators and assist them to interact with surroundings Smart operators have been proposed for complex human-machine-product interactions by integrating AR contents and intelligent tutoring systems Cloud-assisted industrial robots perform tasks with the capacity of interaction and negotiation. Cyber-physical system (CPS) and pervasive technologies are applied to improve the adaptivity of the machine behavior to the working conditions and the specific workers' skills, tasks, and cognitive-physical abilities are improved for aging workers An RFID-enabled positioning system in an automated guided vehicle for logistics automation A method on the perspective of both macro and micro level is developed for correctness analysis of cooperative behaviors among industrial devices A approach of facilitating the large-scale online multitask learning and decision-making is developed for operators to perform flexible tasks A multi-agent system with the autonomy can achieve big data-based feedback and coordination to assist the central coordinator A dynamic model for supply chain scheduling to solve simultaneous consideration of both machine structure selection and job assignments An object-oriented workflow language is developed for formalizing processes with the heterogeneous and dynamic environment in CPS The smart workflow is developed by the adoption of automatic identification technologies which can modeling and reengineering business processes Characteristics Auto nom Adap Socia y tivity bility √ √ √ √ √ √ √ √ √ √ √ √ √ √ √ √ √ √ √ √ √ 191 Compared with traditional task execution process, SWP has many unique characteristics, including 192 traceability, value-added, and awareness. However, information communication, adaptive to 193 changes, autonomous actions during task executions have been identified as the necessary 194 requirements of SWP in previous studies (Lu et al.2017; Wang et al. 2016b; Ren et al. 2017; Lee 195 et al., 2009). Based on using simulated or historical data, SWP could achieve autonomy by 196 executing particular tasks when specific requirements are met (Lu et al., 2017). In addition, each 197 smart work package can gain sociability by communicating with its internal elements, as well as 198 other smart work packages (SWPs) to work as a distributed multi-agent system for collaborative 199 working (Ren et al., 2017). Most importantly, SWP must be adaptive and can react flexibly to 200 changes by learning from its own experiences, environment, and interaction with others (Wang et 201 al. 2016b; Lee et al., 2009). Thus, it is believed that the three critical characteristics of SWP are 202 autonomy, adaptivity, and sociability. The potential functions of SWP have also been introduced 203 and assessed in different scenarios including modeling (e.g., the understanding of the 204 interconnections among tasks), monitoring (i.e. the tracking and updating of real-time status), and 205 optimization (i.e. the planning and scheduling of tasks) (Luo et al. 2018; Wan et al. 2018; Zhang 206 et al. 2018). 207 However, it should be noted that SWP and its definition, characteristics, functions , applications, 208 and prospects in the PHP field have not yet been systematically explored for constraints 209 management. Although individual SWP studies have been investigated, they do not provide a 210 systematic view to explore the full potential of SWP, which is a necessity in driving toward a 211 sweeping and interconnected smartness in next-generation PHP practice, particularly in the field 212 of constraints management in PHP. This requires an investigation of the unique and inherent 213 characteristics of SWP from the manufacturing industry and the incorporation of PHP 214 characteristics. 215 3. Definition of SWP 216 In this study, SWP is defined as an approach to decompose the PHP workflows (e.g., technical 217 process) by product breakdown structure (PBS) of building systems, and integrate smartness 218 capabilities, such as visualizing, tracking, sensing, processing, networking, and reasoning into the 219 workflows so that they can be executed autonomously, adapt to changes in their physical context, 220 and interact with the surroundings to enable more resilient process. 221 The core characteristics of SWP, namely, adaptivity, sociability, and autonomy. Physical or 222 functional information, such as shape, dimension, products type, the layout of the work section, 223 work procedure, and positions of aids and resources, are not included because such information is 224 also required in traditional work packaging method. 225 Adaptivity, the most distinct feature of SWP compared with traditional PHP work packaging 226 method, denotes SWP’s ability to have a positive response to change, and learn from their own 227 experiences, environment, and interactions with others. This characteristic is based on the concepts 228 of smart workflows proposed by Wieland et al. (2008), which includes three dimensions, e.g., 229 robustness, flexibility, and resilience (Husdal, 2010). Robustness is the fundamental feature level 230 that the SWP can process. With robustness, SWP can quickly regain stability by accepting goal- 231 directed initiatives when encountering constraints. It can be mainly applied to plan and control 232 primitive tasks, which refer to elemental motion with few steps or short durations. For instance, 233 the crane operator with the help of SWP can regain stable reaching, grasping, picking up, moving, 234 and eye travel in the lift operations when encountering static constraint such as obstacles. 235 Flexibility enables SWP to react to the foreseeable changes in a pre-planned manner. It is beneficial 236 for guarding tasks execution against threshold-breaking or exceeding a pre-programmed tolerance 237 range, and the SWP in this context primarily involves composite tasks such as to measure, connect, 238 navigate, select, align, record, and report. For example, SWP can help crane operators measure the 239 distance and report the parallax error when other tower cranes are approaching. Resilience is a 240 high-level adaptivity that facilitates SWP to survive unforeseeable changes (that have severe and 241 enduring impacts) in a dynamic replanning manner. The SWP tasks in this context include 242 operation-specific tasks such as assembly, examining workflow, buffer layout, equipment path 243 planning, and monitoring. For example, when an emergency occurs, SWP with resilience can offer 244 assembly guidance and perform the optimized working path planning by cross-validating the real- 245 time progress with as-planned workflow. Presently, SWP adaptivity can be achieved by advanced 246 optimization approaches when making full use of the information collected from the sensing and 247 tracking technologies. 248 Sociability ensures that SWP can communicate with the surroundings (e.g., other smart work 249 packages (SWPs), human/machine/products in SWPs). The communication can happen at pull, 250 push, or mixed modes. The pull mode occurs upon demand. For instance, the 251 deliverables/information, such as prefabricated products from the transportation driver, are 252 provided when requested by the SWP of the expeditor. In the push mode, SWP actively tracks and 253 updates the information and issues alerts at regular intervals or when an emergency occurs. For 254 example, the project manager of the SWP can obtain the traceability and visibility of the 255 prefabricated products in a real-time manner to ensure its Just-in-time delivery. The mixed mode 256 combines the pull and push to request and deliver information in a peer-to-peer manner. Apart 257 from the three interaction modes of SWP, there are four relationships between SWPs, namely, 258 composition, interface realization, inheritance, and dependency, which can enhance the sociability 259 of SWP in handling the modular products/processes in PHP (Ramaji et al., 2016). Composition 260 refers to the relationship of one SWP and its relevant SWPs. For instance, the work package of 261 schedule management usually includes planning, progress checking, monitoring, and risk control. 262 Interface realization refers to a group of work packages which support or rely on the behavior that 263 is defined in an interface. Inheritance exists between a parent smart work package and its 264 succeeding sub-SWPs. Dependency is the most popular relationship where the downstream SWPs 265 are dependent on the upstream SWPs. To achieve the sociability of SWP, there are many 266 communication and networking technologies to enhance the awareness of SWP such as 267 active/passive RFID, ultrawideband (UWB), ZigBee, electromagnetic, Bluetooth, ultrasound, 268 infrared (IR) proximity, Wi-Fi, near-field communication (NFC), laser, conventional radio 269 frequency (RF) timing, wireless local area network (WLAN), received signal strength (RSS), and 270 assisted GPS (A-GPS) (Niu et al. 2016; Zhang and Hammad, 2011). 271 Autonomy proposed in this study is based on the concept of SCOs (Niu et al., 2016). It refers to 272 the capability of intelligent resources (e.g., machinery/tools/devices) in SWP to achieve autonomy 273 through a pre-programmed method of decision making. There are three types of autonomy, 274 including proactive autonomy, passive autonomy, and a mixed mode. Proactive autonomy aims to 275 act in advance of a future situation. For instance, the autonomous crane tower can generate a lift 276 plan in accordance with the dynamic construction environment. It can sense and monitor the 277 dynamic constraints in the environment to predict and execute the plan in advance, without human 278 interventions. Passive autonomy, on the other hand, can only perform instant reaction by a 279 triggering mechanism, particularly triggered by the emergent situation due to the delays of 280 personnel reactions. For example, the anti-heat stress uniform encapsulated in the SWP can issue 281 an alert to the workers and help to reduce heat and humidity when they exceed a certain threshold 282 (Yi et al., 2016). The mixed mode of autonomy may execute complex tasks involving multi- 283 autonomy stages that can both control activities without intervention and act in a preset manner. 284 For instance, the path planning in SWP of a crane operator can firstly be pre-programmed with 285 optimal paths and collisions can be detected in the operation process with the dynamic autonomy. 286 The three core characteristics of SWP are interrelated. Each subclass of the adaptivity, sociability, 287 and autonomy is not a bijection. Instead, various subclasses of characteristics can be integrated to 288 address specific constraints. In more complicated scenarios, it is also possible that the integration 289 of characteristics that are more advanced than these three features is needed. However, this is 290 currently beyond the scope of this study. 291 4. Research Method 292 The development of the conceptual framework started with the definition of the SWP after a 293 comprehensive review of the work packaging method, constraints management, and the smartness 294 concept. Afterward, a draft paradigm, as shown in Figure 1, was proposed as the backbone of the 295 framework. Constraint modeling is included in the SWP to facilitate the identification and 296 interrelationship mapping of the constraints at the activity level (e.g., on-site assembly process). 297 Then, the most influential constraint at the activity level to the goal (e.g., schedule performance) 298 is isolated for further improvement, and this constraint often also contains many constraints at the 299 task level (i.e., specific onsite operational activities). The constraints optimization service in SWP 300 can help develop the optimal task executions by optimizing the constraints at the task level. 301 Tracking, updating, and predicting the statuses of the constraints at the task level are also included 302 in the framework. Constraint at the activity level (e.g., on-site assembly process ) Constraint at the task level (e.g., crane path planning) Identify the critical one Constraint Modeling Constraint Optimization Identify the critical one Constraint Monitoring Fig. 1. The paradigm of SWP-enabled constraints management 303 304 In addition, a layered system model, as the functional structure of SWP in PHP, was also proposed 305 to instantiate the conceptual framework. Its development is based on previous studies on IoT- 306 enabled BIM platforms for PHP (e.g., Li et al. 2017a; Li et al. 2018b), in order to take advantage 307 of both smart BIM platforms and smart construction objects in PHP. 308 Subsequently, the proposed framework and the layered system model were examined and finalized 309 by 14 PHP industry professionals, who were the primary stakeholders of PHP in Hong Kong. All 310 14 experts investigated the framework and provided their comments on the potential application 311 scenarios and functions based on their expertise. As shown in Table 2, the invited professionals 312 included stakeholders from the client, contractor, manufacturer, transportation company, and 313 consultancy. All industry professionals had more than 10 years of experience in the development, 314 operation, and management of PHP projects and related technologies. It is therefore expected that 315 these PHP professionals did provide an unbiased and constructive assessment of the framework. 316 Table 2 The Background of the 14 Interviewees and Their Contribution 317 No. 1 2 Organization (Stakeholder) Housing Authority (Client) 3 Expertise Years of Experience Main Contribution Construction Management 20+ Constraints Identification Supply Chain Management 20+ Applications of Functions Lean Construction 20+ Example Scenarios Production Management 20+ Constraints Monitoring Construction Management 15+ BIM 10+ Construction Management 15+ Housing Society (Client) 4 5 6 7 Gammon Construction (Contractor) Aggressive (Contractor) Constraints Modeling in On-site Assembly Process The system model of SWP Constraints Optimization in Onsite Assembly Process 8 BIM 10+ The function model of SWP 9 Prefabrication Production 10+ Production Breakdown System 10 Process Operations 10+ Constraints Optimization in Production Process 11 Supply Chain Management 15+ Constraints Modeling in Supply Chain Process Logistics and Positioning Technologies 10+ Constraints Monitoring in the Logistics Process WHS (Manufacturer) MDM (Logistics) 12 318 13 CIC (Consultancy) Lean Construction 20+ Framework 14 TSL (IT Consultancy) IoTs Solutions 15+ Properties of SWP 319 320 In order to validate the proposed framework of SWP-CM, a laboratory test was also conducted by 321 using a simulation game (named RBL-PHP, RFID/BIM/Lean-PHP, a role-playing game) 322 developed by the authors (Li et al., 2017a). The following questions were raised: 323 • Can the constraints in PHP workflow be intelligently identified, improved, and monitored? 324 • Can the framework reduce project duration to improve the reliability of PHP workflow? 325 • Can productivity be increased in the implementation of this framework? 326 The aim of the game was to simulate a real-world PHP environment by building LegoTM houses. 327 The task goals were to construct four buildings with the shortest duration, the highest accuracy, 328 and the maximum percentage of the plan complete (PPC). Figure 2 shows the roles and the number 329 of people needed in this simulation game. All the 32 volunteers were postgraduate students with 330 limited knowledge of SWP and constraints management, and ten of them had more than three years 331 of working experience in the construction industry. Such an arrangement can help collect 332 comments, suggestions, and insights from the perspectives of both academic scholars and industry 333 practitioners. The volunteers were divided into two groups, who played in two separate rounds. 334 The first round was related to the use of traditional planning and control (without SWP techniques), 335 and the second round was related to the implementation of SWP-CM. These two rounds were then 336 comparatively analyzed to demonstrate the benefits and differences in implementing the proposed 337 framework. In order to reduce the influence by learning curve issues, there was a briefing session 338 for both rounds, and the participants were also instructed to play before the game. 339 340 5. The Framework of SWP-enabled Constraints Management (SWP-CM) 341 This section outlines the framework of SWP-CM, which aims to improve the workflow of PHP. 342 After the review from selected industry experts, the client of HK Housing Society with the 343 background of Lean Construction agreed that there are two levels of constraints in the PHP process, 344 namely activity-level and task-level constraints, but he also pointed out that the framework should 345 not only reflect the concurrent and continuous improvements of constraints from a perspective of 346 Lean principles but also clarifies the process to the goal by identifying the critical chain of the 347 constraints based on the theory of constraints. An expert from the contractor emphasized the 348 alignment of work packaging stage among activity-level planning, task-level planning, and task 349 executions. In addition, the expert from CIC highlighted the implementation of the three 350 constraints management steps in this framework could help analyze the constraints and their 351 interrelationships systematically in the whole activity process, along with providing the executable 352 plan to remove the constraint at a more detailed level. However, the three steps of the framework 353 should be well-defined in SWP. The IT consultancy mentioned the capabilities of IoT and 354 emerging technology solutions and the integration of these technologies into the framework. A 355 project manager from the client emphasized that the fusion of SWP and constraints management 356 under a clear application scenario should be well considered. SWP1 SWP6 SWP10 SWP7 SWP14 SWP13 SWP2 SWP3 SWP8 SWP4 SWP5 SWP9 SWP11 Fig. 2 Roles and layout of the simulation game SWP12 357 Figure 3 presents the final version of the SWP-CM framework. The work packaging method with 358 lean principles is designed as the basis to outline the workflow of the activity or task execution in 359 PHP. In addition, the framework shows the three core modules of constraints management, 360 followed by the detailed process of SWP-CM. 361 362 To achieve the successful implementation of this framework, three functions, including constraints 363 modeling, constraints optimization, and constraints monitoring must be well combined with the 364 core characteristics of SWP for constraints management in PHP. 365 5.1 Constraints Modeling 366 Constraint modeling is a critical function with the sociability to allow a thorough understanding 367 of interconnections among tasks or activities. There are three steps within this function. The first 368 step is the constraints identification. The traditional process for constraints identification is static 369 and usually executed once. The SWP can enhance this step in a passive autonomy manner by pre- 370 programming the list of constraints and their classification with an open-data integration approach 371 for constraints instantiation. Although each PHP project is unique, they share some similar types 372 of constraints at the operational level (Li et al., 2018a), and it is possible to develop a database for 373 organizing the potentially significant amount of constraints. Table 3 demonstrates the one example 374 of constraints classification in the PHP process, which was sourced from the literature review and 375 the on-site survey. These constraints are classified into manufacturing, logistics, and site 376 constraints. Constraints such as incomplete design drawings/BIM models, approvals, and 377 specifications are manufacturing constraints, which restrict the subsequent activities in logistics 378 and on-site assembly. Logistics constraints contain limited weight and height for vehicles on the 379 road, unavailable production schedule, and transportation schedule. Without JIT deliveries, the site Fig. 3 The proposed framework of SWP-enabled constraints management in PHP 380 buffer may be congested, or underutilized and on-site assembly cannot be efficiently executed. 381 Site constraints include inadequate buffer and workspace, unavailable and unassigned labor 382 resources, lack of collision-free crane path planning, lack of optimal installation sequence, and 383 adverse weather conditions. The reason for this classification is that Manufacturing, logistics, and 384 on-site assembly are the most critical stages in PHP, which can facilitate crews to identify the 385 constraints in their stages. Once the list is embedded into the SWP, a set of pre-defined constraints 386 and their relationships will be available for critical constraints identification. 387 Table 3 List of Constraints and Their Classification 388 Classification Manufacturing Constraints Availability of mould, machinery, storage space, approvals, drawings, BIM models, specifications Adverse weather conditions; unavailable production and transportation schedule; Logistics bad conditions of transportation vehicle and route; road/vehicle limitation in weight and height; Lack of real-time vehicle location; Lack of optimal transportation route Availability of prefabricated products and temporary structures; safety &occupational health training; workspace; buffer space On-site Assembly Availability of labor, shop drawings, instructions, quality, inspection hold-points, transportation planning, safety checkpoints, installation sequence, crane lift and place location, collision-free crane path planning; Adverse weather conditions 389 390 391 The second step is the constraints relationship mapping. In real PHP projects, constraints are 392 usually not independent and may have dynamic interrelationships. As such, a thorough 393 understanding of these relationships is necessary. Figure 4 shows a simple example that includes 394 only one crew with SWP in each selected trade (e.g., manufacturing worker, transportation driver, 395 expeditor, buffer foreman, crane operator, installation worker). The constraints for production (e.g., 396 drawings, BIM models, specifications, machinery) can be handled in the SWP of manufacturing 397 worker. The development of SWP for expeditor needs to rely on well-satisfied constraints of 398 vehicle locations, production, and transportation schedule in SWP of transportation driver. 399 Therefore, any failure of constraints improvement in each SWP may lead to subsequent SWP delay 400 in task executions. The control theory-based system dynamics (SD) model have the capacity to 401 analyze the interactions (e.g., casual loop) and structures (e.g., stock and flow) of the project 402 environments due to their perfect representation of feedback effects. SD models are primarily 403 linked to strategic level context, such as the satisfaction level of the tasks, level of worker fatigue, 404 level of worker skill. The Discrete Event Simulation (DES) can simulate sequential operation 405 details and offer detailed information for execution. Taking the on-site assembly process as an 406 example, the DES model may include detailed information such as the capacity and number of 407 project resources, the duration of on-site assembly tasks, and the lifting distance of the crane tower. 408 Thus, the hybrid SD-DES model can be an alternative to be incorporated into SWP to facilitate the 409 constraints relationship mapping. The last step is the constraints scenario analysis, which can be 410 presented in the interface of SWP for both project managers and workers to show the different 411 simulation results on the schedule performance by evaluating the influence of different critical 412 constraints. The most influential one will be selected for further optimization and monitoring. 413 Fig. 4 An example of constraint relationship mapping 414 5.2 Constraints Optimization 415 5.3 Constraints Monitoring 416 In PHP projects, the latest constraints information is essential for the superintendent or workers to 417 check the progress and issue constraint-free SWPs. As such, real-time constraints monitoring is 418 needed. There are three processes within the function of constraints monitoring. The first process 419 is constraints tracking, which focuses on tracking each individual constraint. For tracking purposes, 420 a mixed type of autonomy is preferred. For instance, the availability of prefabricated products can 421 be tracked by both active and passive RFID (or IoT systems) and visualized in the BIM as the 422 interface of SWP (Li et al., 2018b). The second process is constraints status updating, which 423 concentrates on computing the maturity of a task. The maturity index can be used to support short- 424 term decision-making in a mixed type of sociability. As shown in Fig.5, Fig.6, and Fig.7, it is the 425 interface of a smart work package for the site expeditor. Firstly, it can enable site expeditor with 426 the ability to update the status of the prefabricated products’ locations in a real-time manner. Fig.5 427 shows each prefabricated product with their ID, status (produced, arrived, or erected), time, 428 latitude, and longitude measured by GPS. At the same time, the digital twins (e.g., BIM models) 429 of smart objects (e.g., prefabricated products mounted with RFID and GPS) can be visualized at 430 regular intervals or via ad-hoc networking on the expeditor interface of SWP for monitoring (as 431 shown in Fig.6). Additionally, it can display locations of trucks in the google map and reveals the 432 task maturity of logistics associated smart work packages for each truck and driver by three status 433 (truck loading, cross-border, arrived) in Fig.7. This can guarantee the prefabricated products being 434 transported to achieve JIT delivery, i.e., the pull perspective. The final process within this function 435 is constraints predicting and alerting. The constraints alerting aims to warn the variations by 436 comparing as-planned constraints improvement plan and real-time constraints status. Historical 437 variation can be used to train and predict the next variation in a robust manner. 438 439 440 441 6. The Functional Structure of SWP: Layered System Model 442 To achieve the characteristics and functions of SWP, a three-layered system is proposed (See 443 Fig.8). 444 445 The context provisioning layer (CPL) is capable of managing the context information of PHP 446 processes, which is often referred as both physical and functional information (e.g., dimension, 447 quantity, specifications, location, resources status). For CPL, BIM platforms can be adopted 448 because it has proven to be an effective digital platform to offer users with the ability to generate, 449 integrate, analyze, simulate, visualize and manage the physical and functional information of a 450 facility (Li et al., 2017b). In addition, it can also support the development of various context-aware 451 applications through application programming interfaces (APIs). The BIM models can also be 452 used to integrate context from multiple sources (e.g., dynamic sensor data, smart construction 453 objects, internet of things) for value-added services. The BIM models can be utilized to break 454 down the design into many units, and each unit comprises various materials, components, and 455 modules. All the prefabricated products within a unit can be grouped into a product work package 456 (PWP), which is in accordance with the product breakdown structure of building systems. The ID Latitude Longitude Method Status Time B5-37F-01-TX8 22.414649 113.975509 GPS Arrived 21/04/2016 10:30:20 GMT +0800 (HKT) B5-37F-03-TX4 22.414266 113.975537 GPS Arrived 21/04/2016 10:30:28 GMT +0800 (HKT) B5-37F-03-TX8 22.414235 113.975537 GPS Arrived 21/04/2016 10:30:40 GMT +0800 (HKT) B5-37F-03-TX9 22.414519 113.975143 GPS Arrived 21/04/201610:30:58 GMT +0800 (HKT) B5-37F-04-TX8r 22.414175 113.975421 GPS Arrived 21/04/2016 10:31:14 GMT +0800 (HKT) B5-37F-04-TX9r 22.414276 113.975421 GPS Arrived 21/04/2016 10:31:29 GMT +0800 (HKT) B5-37F-05-TX8 22.414576 113.975485 GPS Arrived 21/04/2016 10:31:48 GMT +0800 (HKT) B5-37F-05-TX9A 22.414602 113.975483 GPS Arrived 21/04/2016 10:31:59 GMT +0800 (HKT) B5-37F-06-TX2r 22.414641 113.975514 GPS Arrived 21/04/2016 10:32:10 GMT +0800 (HKT) Fig. 5 Location status of each prefabricated product (Status Tracking) Fig. 6 Visualized status of each prefabricated product in BIM (Status Monitoring) Truck Loading Cross-border Clearance Driver1_Truck_No.KX9038 100% Driver2_Truck_No.WJ6809 100% 67% Driver3_Truck_No.LX5537 Driver4_Truck_No.TY0842 Arrived 33% Fig. 7 Visualized location status of each truck and the task maturity of each driver (Status Updating) SWPL SWPi-1 Issue an SWP Execution of SWP Wait for Completion SWPi+1 Context Integration Processes Functions Query resources Transport Complete Event Query algorithms Query location Domain specific CIPs Query sequence CIL Location-based Workflow Engine Integration of PWPs with Location-based workflow Context Query CPL Context Insert Context Manipulation Core CIPs Context Event BIM Platform: Assignment of components to product work packages (PWPs) Fig. 8 Layered System Model of SWP 457 PWP will then be decomposed into SWPs by integrating the context of the workflows (process), 458 work faces (location), duration, and resources. 459 The context integration layer (CIL) adopts the output of CPL to accommodate information, 460 algorithms, and functions into more advanced representations and provide domain-specific 461 functions needed by SWPs. Compared with CPL, there is no off-the-shelf system for CIL. The 462 primary contribution of this model is to present the concept of how to design this layer. There are 463 two context integration processes (CIPs) for CIL, namely (1) Core CIPs, and (2) Domain-specific 464 CIPs. Within a location-based workflow engine, the former can help map the physical products, 465 data, and services into the specific location-based workface to integrate the necessary elements for 466 work packages, while the latter can help workers with different domain knowledge extract well- 467 formatted work packages from Core CIPs and access different functions. In the core CIPs, the BIM 468 model can be decomposed into various prefabricated products with both physical and functional 469 information, which can form different product work packages (PWPs). Then, these PWPs can be 470 integrated into the workflow of the PHP process (e.g., on-site assembly process). At this moment, 471 the process-oriented information, e.g., the location of workface, technical procedure, required 472 resources, can be integrated with PWPs to generate the work packages by introducing advanced 473 algorithms (e.g., partitioning algorithms). The integration of PWPs with workflow by CIPs serves 474 an autonomous pattern. A Core CIP receives a call from the workflow (a higher-ranking Core CIP 475 of upstream SWPs) and remodels the request to the required format of the service including context 476 query, insert, manipulation, and event. Context queries facilitate the query to be synchronized with 477 context information, e.g., with a query language. The query result can serve as a variable to be 478 injected into the complex workflow. If the query language allows data manipulation, a workflow 479 can enable the function of context insert and change. The second process is related to domain- 480 specific CIPs and can offer context information at various semantical levels for SWPs. The 481 domain-specific CIPs include two primary functions: one is to merge specific functional elements 482 to the well-formatted work packages from core CIPs to form SWPs; the other is to simplify the 483 interfaces (e.g., web service interface) of SWPs for accessing their functionality. 484 Finally, the SWP layer (SWPL) can not only issue a smart work package with mobile, wearable, 485 and executable capacity but also provide a platform to interact with other SWPs. In addition, any 486 execution failure can trigger the dynamic re-planning function to provide more adaptive SWP. The 487 experts also evaluate the proposed layered system model by their expertise and project experience, 488 and the comments are summarized as follows: “This functional structure of SWP fully utilizes the 489 capabilities of existing BIM platforms and smart construction objects to help equip the workers 490 with more value-added information and make them more skillful on task executions.” (senior IoTs 491 engineer, TSL) “It is feasible to embed this layered system model into the service-oriented 492 architecture of the previous project ‘IoT-enabled BIM Platforms for Prefabrication Housing 493 Production.’” (senior BIM system architect, Gammon Construction) 494 7. Validation 495 7.1 Validation Design 496 A simulation game following the real processes of PHP projects (e.g., a Subsidized Sale Flats 497 project owned by the Hong Kong Housing Society and locates at 48 Chui Ling Road, Tseung 498 Kwan O Area 73A) is conducted through a workshop to assess the validity of the proposed 499 framework. According to the role setting and the proposed framework of SWP-CM, 14 SWPs were 500 developed for the simulation game (See Figure 2 and Table 4). There are three connected scenarios 501 (manufacturing, logistics, and on-site assembly) in this game. A process map was provided to the 502 participants to understand the simulation game. In this study, 13 constraints, including lack of 503 approvals from site manager, design drawings, BIM models, specifications, tools, production 504 schedule, transportation schedule, prefabricated products (e.g., material, components, modules, 505 units), buffer space, assembly instructions, quality and inspection hold-points, crane lift and place 506 location, and vehicle limitation in weight and height, were included. If the project team cannot 507 improve these constraints in an efficient manner, the game may suffer delay. 508 Table 4 Trade-associated SWP 509 SWP_No. 1 2 3 4 5 Trade Manufacturing Worker Manufacturing Worker Manufacturing Worker Manufacturing Worker Manufacturing Worker SWP_No. 6 7 8 9 10 Trade Plant Manager Project Manager Truck Driver Logistics Manager Expeditor SWP_No. 11 12 13 14 Trade Crane Operator Site Worker Site Manager Buffer Foreman 510 511 512 The first round of the game focused on the SWP-CM framework. The constraints identification 513 process was conducted to synchronize the constraints list and the constraint relationship map to 514 the SWP, which could be accessed by each participant through mobile devices. This process was 515 achieved at the beginning of the game in the social network analysis (SNA) service of SWP, which 516 included three primary steps: (1) The participants registered in the SNA service of their own SWP 517 and accessed the full list of constraints; (2) The participants scored and evaluated the constraints 518 interrelationships; (3) The participants visualized the constraint network and identified critical 519 constraints and constraint interactions. After the identification, a hybrid system dynamic (SD)- 520 discrete event simulation (DES) model service was adopted to assess and simulate the potential 521 effect of the identified constraints on the schedule performance. DES was adopted to measure the 522 operation level of game and SD was related to the strategic level consideration, including resource 523 availability, operation efficiency, and schedule performance. Finally, the constraints analysis 524 results were also demonstrated to participants by embedding the results in specific SWP buttons. 525 As shown in Figure 9, when clicking “Expeditor_SWP,” the expeditor could find all related 526 constraints and other interactional SWPs. After clicking the specific constraint in each SWP, the 527 simulation results can be presented. Apart from the constraints modeling, the detailed task 528 execution plans for improving each constraint are also presented. Lean principles, such as pull 529 methods, Just in time delivery, and standardized work, served as the optimization strategies in this 530 simulation game. For instance, the pull method can be used to improve the constraints “lack of 531 production schedule” in the SWP_11 (See Figure 9) for expediting the production process. 532 Furthermore, the status of each constraint was also tracked and visualized through the use of RFID 533 tracking technology and BIM visualization interface (see 10.2 “prefabricated products traceability” 534 in Figure 9). With SWP-CM implementation, Group A was able to detect and analyze all 535 constraints in the first 9 minutes and adopt relevant optimization strategies. The first round took 536 35 minutes, and the performance of Group A was evaluated by the percentage of plan complete 537 (PPC), productivity index, and extra cost. The definition of these three indicators and their 538 calculations are shown in Table 5. 539 540 Step No.1 13. Site Manager_SWP Current Production Status 14. Buffer Foreman_SWP Step No.2 Step No.5 Step No.3 Step No.4 Step No.7 Step No.6 8. Transportation Driver_SWP Unit Red1 Red2 Red3 Red4 Maturity 30% 20% 10% 10% Unit White1 White2 White3 White4 Maturity 50% 40% 30% 10% Unit Blue1 Blue2 Blue3 Blue4 Maturity 100% 70% 50% 30% Unit Maturity Yellow1 90% Yellow2 80% Yellow3 70% Yellow4 60% Step No.8 10. Expeditor_SWP If 10.1_delay = 1min ; Schedule _ delay = 2 min If 10.3_delay = 1min ; Schedule _ delay = 3min 10. 1 Production Schedule 10. 3 Transportation Schedule Production If 10.2_delay = 1min ; Schedule _ delay = 4min 10. 2 Prefabricated Products Traceability Red 1 Red 2 Red 3 Red 4 Delivery to Site Arrive at Buffer 100% 75% 50% 25% Fig. 9 Detailed constraints improvement in Expeditor_SWP Assembly 541 Table 5 The description of indicators for validation 542 Indicator PPC Extra Cost Description To measure the actual completion at the end of each time interval. (1) time interval = 9 min in this study. (2) A total number of units to be constructed = 20. This may result from the overly produced units that are transported to the construction site, the defective units that need rework, and the manufactured-inprocess (MIP) units that cause delay. Formula Remark PPC = Qa / Qt Where Qa = the number of assembled units, Qt = the total number of units (20). The cost of each unit can be found in the authors’ previous work (Li et al., 2017a) The cost of each component contains the cost of material, labor, equipment, and transportation. a. Pm = (Qp – Qd1) / (Tf1 – Ts1) Productivity Index This is a measurement of the ability to manufacture, transport, and assemble. b. Pl = (Ql – Qd2) / (Tf2 – Ts2) c. Pa = (Qa – Qd3) / (Tf3 – Ts3) where Pm = the productivity index of manufacturing; Qp = number of produced units in the plant; Qd1 = number of defective units in the plant; Tf1 = finish time of the production of the last unit; and D1 = duration from Ts1 to Tf1 and D1 = Tf1 – Ts1. where Pl = the productivity index of logistics; Ql = number of transported units; Qd2 = number of defective units in the logistics; Tf2 = finish time of the transportation of the last unit; and D2 = duration from Ts2 to Tf2 and D2 = Tf2 – Ts2. where Pa = the productivity index of onsite assembly; Qd3 = number of defective units in the assembly process; Tf3 = finish time of the assembly of the last unit; and D3 = duration from Ts3 to Tf3 and D3 = Tf3 – Ts3. 543 544 545 The second round game focused on the traditional constraints improvement method. The following 546 changes were made, while other conditions remained the same. 547 (1) Constraints modeling, including the relationship map and analysis results, were not provided 548 to Group B. Based on the inputs of the 14 industry professionals, constraints identification, 549 relationship mapping, and analysis were conducted informally on the basis of experience. 550 (2) Constraints optimization strategies were only developed when the constraints happened. The 551 participants could discuss optimal solution strategies in a meeting when constraints occurred. 552 (3) The players were not allowed to directly monitor others who have geographical barriers in real 553 situations. In this simulation, they can arrange regular coordination meetings to report their own 554 progress. 555 As there was no implementation of SWP-CM, the 13 constraints had not been timely identified 556 until the second 9-minute interval. The game suffered delay due to the late removal of the 557 constraints (e.g., shortage of tools and prefabricated products) and the performance was also 558 measured by the same indicators. 559 7.2 Validation Results 560 The results are shown in Tables 6-8, respectively. Table 6 demonstrates the actual duration and 561 the PPC values of the two rounds. A total of 35 min was recorded in the first round while the 562 second round took 45 min, which suggests that 22.2% reduction in project duration was achieved 563 through the implementation of SWP-CM. The main underlying reason was the late identification 564 and improvement of the constraints in the second round, and participants spent more time 565 understanding the constraints and identifying optimization strategies. Table 7 shows the results of 566 the simulation game at extra cost. An extra cost of $7460 was recorded in the second round while 567 there was no extra cost in the first round. In the second round, as the push system without 568 constraints monitoring was adopted, two additional units were produced, and one unit was 569 manufacturing-in-process (MIP). Table 8 shows the productivity index of the two rounds. The 570 productivity is significantly improved in all three phases, including manufacturing (Pm : 0.53 → 571 0.67; 26% increase), logistics (Pl : 0.88 → 1; 14% increase), and on-site assembly (Pa : 0.49 → 572 0.65; 33% increase). Efficient information sharing and communication in the first round 573 demonstrated the effectiveness of the real-time constraints modeling, optimization, and monitoring, 574 which can be considered as the main contribution to the increase in productivity. 575 Table 6 The Percentage of Plan Complete 576 Round Actual Duration (min) PPC at the end of the first 9 min (%) PPC at the end of the second 9 min (%) PPC at the end of the third 9 min (%) PPC at the end of the fourth 9 min (%) PPC at the end of the fifth 9 min (%) Round 1 35 20 45 75 100 - Round 2 45 10 30 55 75 100 577 Table 7 The Extra Cost in the Simulation Game 578 Round Overproduced units (Qty) Defective Units (Qty) R W R 1 1 B Y W B Y Total Extra Cost ($) 0 1 7460 MIP(Qty) Y R W B Round 1 Round 2 1 1 Note: R = “Red Unit”, W=“White Unit”, B=“Black Unit”, Y=“Yellow Unit” 579 580 Table 8 The Productivity Index in the Simulation Game 581 Round Qp Qd1 Ql Qd2 Qa Qd3 D1(min) D2(min) D3(min) Pm Pl Pa Round 1 20 0 20 0 20 0 30 20 31 0.67 1 0.65 Round 2 22 1 22 0 20 1 39 25 40 0.54 0.88 0.49 582 583 In summary, the round with SWP-CM outperforms the traditional round. The results also answer 584 the previously raised questions with the following evidence: (1) Several intelligent techniques (e.g., 585 SNA, DES, SD, Lean tools, BIM) have been used in constraints modeling, optimizing, and 586 monitoring to achieve the certain level of sociability, adaptivity, and autonomy in the SWP-CM 587 round; (2) The duration was reduced by 22.2% in in the SWP-CM round and $7460 extra cost 588 occurred in the traditional round; (3) The productivity in the phases of manufacturing, logistics, 589 and on-site assembly was increased with 26%, 14%, and 33%, respectively. 590 8. Discussion 591 Constraints management in modern PHP projects is essential because PHP processes are separated 592 into different stages. Existing approaches to constraints management have several shortcomings, 593 including low transparency of constraints status, and non-optimal or inflexible constraints 594 improvement planning (Wang et al. 2016a). The previous manual and people-centric approaches 595 in constraints management disregard the potential of IT to accurately, timely, and agilely in 596 managing constraints, thus enabling the reliable workflow in PHP scenarios. With smart 597 characteristics, including adaptivity, sociability, and autonomy, SWP can strengthen constraints 598 modeling, monitoring, and even optimization. Accordingly, SWP can improve human deficiencies 599 or skills in tasks execution to save time and cost. SWP can identify and analyze the latest 600 constraints in a pull or push manner, provide optimal constraints improvement planning at different 601 levels such as robustness, flexibility, resilience, and track, update, and predict the constraints status 602 autonomously. 603 SWP provides an immense opportunity to improve workflow management in the global 604 modular/prefabricated construction industry. SWP can significantly enhance the power of object- 605 oriented BIM, which has been broadly recognized as a potential of integrating physical objects of 606 product-oriented PHP and informational components to form situation-integrated analytical 607 systems which can respond intelligently to the dynamic changes of real-world scenarios and offer 608 data-oriented lean solutions (Li et al. 2017b). Current BIM models are mostly created in an as- 609 designed condition, with updates in the subsequent stages including construction and maintenance. 610 To make BIM a handy information hub in tasks execution with data-oriented lean solutions, as- 611 built information is urgently needed to timely exchange with BIM. Presently, as-built data updates 612 are primarily based on manual site survey or fragmented information technologies adoptions, 613 which are time-consuming, error-prone, and non-value added information (Shrestha and Behzadan, 614 2018). To some extent, BIM development for physical project execution has come to a bottleneck 615 with as-built information being synchronizing between BIM and tasks execution in a real-time and 616 value-added manner to support constraints management. SWP can be adopted to bridge the value- 617 added information gap between BIM and information technologies supported objects (e.g., smart 618 PHP objects). The sociability of SWP means that they can interact with other SWPs or synchronize 619 as-built information with BIM in a pull or push manner, and the adaptivity of SWP can make them 620 respond to changes in a robust, flexible and resilient manner. The characteristic of autonomy 621 enables SWP to respond in a proactive or passive manner. 622 Given the capacity of SWP to interact with other platforms, SWP can also benefit from the 623 development of the Internet of Things (IoT), an emerging paradigm that has attracted considerable 624 attention in the lifecycle of PHP (Li et al., 2018b), In the IoT paradigm, the constraints status can 625 be connected at any time and anywhere. The gateway, an IoT-enabled industrial computer, can 626 provide a communication link between physical sensors and SWPs. Thus, IoT can enable the 627 SWPs to be a loosely coupled, decentralized, multi-agent system. The adaptivity held by SWP is 628 a core property in the IoT ecosystem, as the flexible and resilient actions can make the planning 629 and control of constraints more dynamic. With the characteristic of autonomy, SWP can connect 630 with and handle the autonomous objects (e.g., vehicle, crane, robotics, 3D printer) based on 631 specific protocols, e.g., a fill-up based trigger (Wu et al., 2016). Once the smart workflow is 632 established, information sensed by each autonomous object can be shared with SWP in a proactive 633 manner. These all contribute to the underpinning philosophy of construction industry 4.0 (Longo 634 et al., 2017). 635 Furthermore, a smart work package can be generated from BIM by decomposing the BIM models 636 and integrating the functional information such as tasks sequence, workflow, resources, location 637 with the decomposed physical information including building systems and prefabricated products. 638 Its information can be pulled out from context provision layer for assisting constraints modeling 639 (e.g., automatic analysis of the topological constraints and their interrelationships), optimization 640 (e.g., visual guidance and interactive representation of the work sequence can be obtained by 641 applying optimal lean solutions), and monitoring (e.g., the resource requests can be evaluated and 642 monitored in a real-time manner). The functions of SWP are developed and integrated into the 643 context integration layer in a specific format (e.g., ifcXML), which can be connected to BIM. Files 644 using the IFC schema can be interoperated on BIM platforms, which facilitates better information 645 sharing and exchange (Lee et al. 2016). SWP also reduces manual operations, including 646 reformating or reinterpreting information (e.g., constraints status) when using BIM, thus 647 eliminating the possibility of the error caused by human intervention during data processing. It is 648 envisaged that the proposed SWP can address the bottleneck that limits BIM expansion and present 649 opportunities to make BIM a genuinely dynamic workflow management system rather than the 650 static model management system. 651 It can be envisaged that SWP will progressively override conventional PHP constraints 652 management to develop into an effective workflow management approach in the future. However, 653 there are still numerous challenges to face. Firstly, from an organizational perspective, there will 654 probably be resistance to diverge from the current constraints management practices in order to 655 embrace smartness. Meanwhile, although SWP can help simplify interface management between 656 tasks/activities carried out by different sub-contractors, the adoption of SWP for constraints 657 management is more challenging in PHP projects with multiple tiers of subcontractors. Secondly, 658 from a technical perspective, the interoperability of SWP will also be a challenge. The smartness 659 of SWP relies on efficient data exchange. Without a universal standard for SWPs, there will be no 660 smartness (though presently SWP can be operated based on BIM interfaces which are interoperated 661 through ifcXML). The PHP industry is also fragmented. No individual can drive the industry 662 toward fully integrated advanced technologies development and adoptions (Niu et al., 2016). The 663 third challenge, from an economic perspective, is the expense of developing and deploying SWP. 664 The PHP industry is comparatively slow-moving to embrace the new wave in the adoption of new 665 technologies, and organizations within the industry would be very sensitive to expand on new 666 technologies. 667 9. Conclusion 668 PHP has fragmented processes, which may generate various constraints in the critical chain of 669 PHP. If the constraints cannot be timely improved, the reliability of workflow may be affected, 670 and schedule delay and cost overrun will occur. The primary contributions of this study to the body 671 of knowledge are threefold. Firstly, Inspired by the theories of work packaging and SCOs, SWP 672 is defined as PHP workflows which are decomposed in accordance with PBS of building systems 673 that are made smart by augmenting with the capacities of visualizing, tracking, sensing, processing, 674 networking, reasoning so that they can be executed autonomously, adapt to changes in their 675 physical context, and interact with surroundings to enable more resilient process. Secondly, 676 equipped with three characteristics sociability, adaptivity, and autonomy, a continuous 677 improvement framework for constraints management with three functions, including constraints 678 modeling, constraints optimization, and constraints monitoring is proposed and illustrated by 679 several examples and scenarios.The rationale and methodology in the framework of SWP-CM can 680 be generalized because the development of the framework does not rely on identifying and 681 removing specific types of constraints.Thirdly, a formal structured SWP representation is proposed 682 by developing a layered system model involving context provisioning layer (CPL), context 683 integration layer (CIL), and smart work packaging layer (SWPL) to realize these three functions. 684 Results from the validation process signify the benefits when implementing the framework of 685 SWP-CM in PHP. 22.2% reduction of project duration was achieved, and no defective units were 686 generated in the round of SWP-CM. Productivity was also improved, particularly in the 687 manufacturing and on-site assembly stage. Thus, it can be concluded that SWP provides enormous 688 opportunities to improve constraints management in PHP, particularly in conjunction with BIM. 689 It can extract the context information (both physical and functional information) of product work 690 packages from CPL (BIM platforms integrating with IoT). It can also insert the value-added as- 691 built information into the BIM platforms in a pull or push manner. SWP can also be combined 692 with the IoT-enabled gateway to act as a loosely coupled, decentralized, multi-agent system to 693 make the status of the constraints be connected at any time and anywhere. 694 However, It should be noted that SWP for constraints management is in the early stage of its 695 development. There are several barriers to the development and implementation. For example, 696 there are technical difficulties related to the integral approach in constraints identification and 697 interrelationship mapping, the efficient algorithms for dynamic re-planning in constraints 698 optimization, and robustness hardware (e.g., autonomous robots, vehicles, cranes) and software 699 (location-based workflow engine, interoperability of connected system) for constraints monitoring. 700 There are also challenges related to technology acceptance, organizational changes, and cost issue. 701 By overcoming these challenges, it is believed that SWP can help establish safer, more adaptive, 702 more proactive, more efficient, and more sustainable PHP workflows. 703 Acknowledgments 704 This research was funded by the Australian Research Council Discovery Project (grant number No. 705 DP180104026), the Linkage Project (grant number No. LP180100222) and the National Key R&D 706 Program of China (No.2016YFC070200504). It was also supported by the Research Institute for 707 Sustainable Urban Development of the Hong Kong Polytechnic University, National Natural 708 Science Foundation of China (No. 71801159), and Natural Science Foundation of Guangdong 709 Province (No. 2018A030310534). 710 Glossary 711 Product Breakdown Structure (PBS): It is a product-oriented planning approach to analyze, 712 document and communicate the outcomes of a project, which offers a comprehensive 713 understanding of the physical deliverables. (Highlights: showing the physical deliverables) 714 Work Breakdown Structure (WBS): It is a deliverable-oriented planning tool to hierarchically 715 decompose the entire scope of work into the combination of product, data, and service that are 716 required in a project. (Highlights: showing the work required to produce deliverables) 717 Advanced Work Package (CWP): It is a planned, executable process that encompasses the work 718 on an EPC project, beginning with initial planning and continuing through detailed design and 719 construction execution. (Highlights: showing the framework of construction execution) 720 Construction Work Package (CWP): It is an executable construction deliverable with the well- 721 defined (e.g., budget and schedule) work scope which cannot overlap with another construction 722 work package. 723 Engineering Work Package (EWP): It is an engineering deliverable with preparation-oriented 724 work scope, which includes drawings, procurement deliverables, specifications, and vendor 725 support to be consistent with the sequence and schedule of CWPs. 726 Installation Work Package (IWP): It is a detailed execution plan that ensures all necessary 727 elements used to complete the scope of the IWP are well organized and delivered before executions 728 to enable workers to perform quality work in a safe, effective and efficient manner. 729 Smart Work Packaging (SWP): It is defined as an approach to decompose the PHP workflows 730 (e.g., technical process) by product breakdown structure (PBS) of building systems that are made 731 smart with augmented capacities of visualizing, tracking, sensing, processing, networking, and 732 reasoning so that they can be executed autonomously, adapt to changes in their physical context, 733 and interact with the surroundings to enable more resilient process. 734 Reference 735 Abuwarda, Z., & Hegazy, T. (2016). Work-Package Planning and Schedule Optimization for Projects with Evolving 736 Constraints. Journal of Computing in Civil Engineering, 30(6), 04016022. 737 Blanco-Novoa, O., Fernández-Caramés, T. M., Fraga-Lamas, P., & Vilar-Montesinos, M. A. (2018). A practical 738 evaluation of commercial industrial augmented reality systems in an industry 4.0 shipyard. IEEE Access, 6, 739 8201-8218. 740 741 742 743 744 745 Boyd, L., & Gupta, M. (2004). Constraints management: what is the theory?. International Journal of Operations & Production Management, 24(4), 350-371. Chi, H. L., Chen, Y. C., Kang, S. C., & Hsieh, S. H. (2012). Development of user interface for teleoperated cranes. Advanced Engineering Informatics, 26(3), 641-652. Giner, P., Cetina, C., Lacuesta, R., & Palacios, G. (2012). Enabling Smart Workflows over Heterogeneous IDSensingTechnologies. Sensors, 12(11), 14914-14936. 746 Gong, P., Teng, Y., Li, X., & Luo, L. (2019). Modeling Constraints for the On-Site Assembly Process of Prefabrication 747 Housing Production: A Social Network Analysis. Sustainability, 11(5), 1387. 748 Goldratt, E. M., & Cox, J. (1984). The Goal, Croton-on-Hudson. NY: North River Press Inc. 749 Hamdi, O. (2013). Advanced Work Packaging from project definition through site execution: driving successful 750 implementation of workforce planning (Doctoral dissertation, The University of Texas at Austin). Retrieved 751 from https://repositories.lib.utexas.edu/handle/2152/21384 752 Housing Authority (2018), Number of Applications and Average Waiting Time for Public Rental Housing, 753 https://www.housingauthority.gov.hk/en/about-us/publications-and-statistics/prh-applications-average- 754 waiting-time/ 755 756 Husdal, J. (2010). A conceptual framework for risk and vulnerability in virtual enterprise networks. In Managing risk in virtual enterprise networks: implementing supply chain principles (pp. 1-27). IGI Global. 757 Ibrahim, Y. M., Lukins, T. C., Zhang, X., Trucco, E., & Kaka, A. P. (2009). Towards automated progress assessment 758 of work package components in construction projects using computer vision. Advanced Engineering 759 Informatics, 23(1), 93-103. 760 Isaac, S., Curreli, M., & Stoliar, Y. (2017). Work packaging with BIM. Automation in Construction, 83, 121-133. 761 Ivanov, D., Dolgui, A., Sokolov, B., Werner, F., & Ivanova, M. (2016). A dynamic model and an algorithm for short- 762 term supply chain scheduling in the smart factory industry 4.0. International Journal of Production Research, 763 54(2), 386-402. 764 Lee, K., Paton, N. W., Sakellariou, R., Deelman, E., Fernandes, A. A., & Mehta, G. (2009). Adaptive workflow 765 processing and execution in Pegasus. Concurrency and Computation: Practice and Experience, 21(16), 766 1965-1981. 767 Lee, Y. C., Eastman, C. M., & Solihin, W. (2016). An ontology-based approach for developing data exchange 768 requirements and model views of building information modeling. Advanced Engineering Informatics, 30(3), 769 354-367. 770 Li, C. Z., Xu, X., Shen, G. Q., Fan, C.,Li, X., Hong, J. (2018a). A model for simulating schedule risks in prefabrication 771 housing production: a case study of six-day cycle assembly activities in Hong Kong. Journal of Cleaner 772 Production, 165, 10 773 Li, C. Z., Xue, F., Li, X., Hong, J., & Shen, G. Q. (2018b). An Internet of Things-enabled BIM platform for on-site 774 assembly services in prefabricated construction. Automation in Construction, 89, 146-161.48-1062. 775 Li, X., Shen, G. Q., Wu, P., Fan, H., Wu, H., & Teng, Y. (2017a). RBL-PHP: Simulation of Lean Construction and 776 Information Technologies for Prefabrication Housing Production. Journal of Management in Engineering, 777 34(2), 04017053. 778 779 780 781 782 Li, X., Wu, P., Shen, G. Q., Wang, X., & Teng, Y. (2017b). Mapping the knowledge domains of Building Information Modeling (BIM): A bibliometric approach. Automation in Construction, 84, 195-206. Li, X., Yi, W., Chi, H. L., Wang, X., & Chan, A. P. (2018c). A critical review of virtual and augmented reality (VR/AR) applications in construction safety. Automation in Construction, 86, 150-162. Li, X., Shen, G. Q., Wu, P., & Yue, T. (2019). Integrating building information modeling and prefabrication housing 783 784 785 production. Automation in Construction, 100, 46-60. Liu, H., Al-Hussein, M., & Lu, M. (2015). BIM-based integrated approach for detailed construction scheduling under resource constraints. Automation in Construction, 53, 29-43. 786 Longo, F., Nicoletti, L., & Padovano, A. (2017). Smart operators in industry 4.0: A human-centred approach to enhance 787 operators’ capabilities and competencies within the new smart factory context. Computers & Industrial 788 Engineering, 113, 144-159. 789 790 791 792 793 794 795 796 797 798 799 800 801 802 Lu, S., Xu, C., Zhong, R. Y., & Wang, L. (2017). An RFID-enabled positioning system in an automated guided vehicle for smart factories. Journal of Manufacturing Systems, 44, 179-190. Luo, Y., Duan, Y., Li, W., Pace, P., & Fortino, G. (2018). Workshop Networks Integration Using Mobile Intelligence in Smart Factories. IEEE Communications Magazine, 56(2), 68-75. Niu, Y., Lu, W., Chen, K., Huang, G. G., & Anumba, C. (2016). Smart construction objects. Journal of Computing in Civil Engineering, 30(4), 04015070. Peruzzini, M., & Pellicciari, M. (2017). A framework to design a human-centered adaptive manufacturing system for aging workers. Advanced Engineering Informatics, 33, 330-349. Ramaji, I. J., & Memari, A. M. (2016). Product architecture model for multistory modular buildings. Journal of Construction Engineering and Management, 142(10), 04016047. Ren, G., Hua, Q., Deng, P., Yang, C., & Zhang, J. (2017). A Multi-Perspective Method for Analysis of Cooperative Behaviors Among Industrial Devices of Smart Factory. IEEE Access, 5, 10882-10891. Seiger, R., Keller, C., Niebling, F., & Schlegel, T. (2015). Modeling complex and flexible processes for smart cyberphysical environments. Journal of Computational Science, 10, 137-148. 803 Shrestha, P., & Behzadan, A. H. (2018). Chaos Theory–Inspired Evolutionary Method to Refine Imperfect Sensor 804 Data for Data-Driven Construction Simulation. Journal of Construction Engineering and Management, 805 144(3), 04018001. 806 807 808 809 810 811 Taghaddos, H., Hermann, U., AbouRizk, S., & Mohamed, Y. (2012). Simulation-based multiagent approach for scheduling modular construction. Journal of Computing in Civil Engineering, 28(2), 263-274. Yi, W., Chan, A. P., Wang, X., & Wang, J. (2016). Development of an early-warning system for site work in hot and humid environments: A case study. Automation in Construction, 62, 101-113. Wan, J., Chen, B., Imran, M., Tao, F., Li, D., Liu, C., & Ahmad, S. (2018). Toward Dynamic Resources Management for IoT-Based Manufacturing. IEEE Communications Magazine, 56(2), 52-59. 812 Wang, J., Shou, W., Wang, X., & Wu, P. (2016a). Developing and evaluating a framework of total constraint 813 management for improving workflow in liquefied natural gas construction. Construction Management and 814 Economics, 34(12), 859-874. 815 Wang, J., Sun, Y., Zhang, W., Thomas, I., Duan, S., & Shi, Y. (2016b). Large-Scale Online Multitask Learning and 816 Decision Making for Flexible Manufacturing. IEEE Transactions on Industrial Informatics, 12(6), 2139- 817 2147. 818 819 Wang, S., Wan, J., Zhang, D., Li, D., & Zhang, C. (2016c). Towards smart factory for industry 4.0: a self-organized multi-agent system with big data-based feedback and coordination. Computer Networks, 101, 158-168. 820 821 Wang, S., Zhang, C., Liu, C., Li, D., & Tang, H. (2017). Cloud-assisted interaction and negotiation of industrial robots for the smart factory. Computers & Electrical Engineering, 63, 66-78. 822 Wieland, M., Kaczmarczyk, P., & Nicklas, D. (2008, March). Context integration for smart workflows. In Pervasive 823 Computing and Communications, 2008. PerCom 2008. Sixth Annual IEEE International Conference on (pp. 824 239-242). IEEE. 825 826 827 828 829 830 Wu P, Wang J, Wang X. A critical review of the use of 3-D printing in the construction industry. Automation in Construction. 2016;68:21-31. Wu P, Song Y, Shou W, Chi H, Chong HY, Sutrisna M. A comprehensive analysis of the credits obtained by LEED 2009 certified green buildings. Renewable and Sustainable Energy Reviews. 2017;68:370-9. Zhang, C., & Hammad, A. (2011). Multiagent approach for real-time collision avoidance and path replanning for cranes. Journal of Computing in Civil Engineering, 26(6), 782-794. 831 Zhang, Y., Wang, W., Du, W., Qian, C., & Yang, H. (2018). Coloured Petri net-based active sensing system of real- 832 time and multi-source manufacturing information for the smart factory. The International Journal of 833 Advanced Manufacturing Technology, 94(9-12), 3427-3439. 834 Zhong, Ray Y., Yi Peng, Fan Xue, Ji Fang, Weiwu Zou, Hao Luo, S. Thomas Ng, Weisheng Lu, Geoffrey QP Shen, 835 and George Q. Huang. "Prefabricated construction enabled by the Internet-of-Things." Automation in 836 Construction 76 (2017): 59-70.