The measurement of the construction productivity of HPCSS and CIPSS was conducted by using the cyclic operation network (CYCLONE) method developed by Halpin and Riggs [
15]. This method is widely used in related research fields. Furthermore, CYCLONE is a discrete-event simulation method that focuses on construction work tasks, and thus, this method is widely applied to model repetitive construction work. In addition, CYCLONE is utilized as a management tool to analyze construction productivity based on the logical connections between work tasks, duration, and resources, and thus, CYCLONE can be used to determine the influence of specific work tasks on the overall construction productivity in conjunction with variations in duration and resources assigned to each work task [
16,
17]. More information on CYCLONE including modeling elements can be found in the study by Halpin and Riggs [
15].
4.1. Simulation Modeling
Using CYCLONE and information with respect to the case study including the work processes (i.e.,
Figure 2) and their precedence relationships, we developed CYCLONE models of CIPSS and HPCSS, as shown in
Figure 3 and
Figure 4, respectively. As shown in the figures, the two simulation models consist of the two following parts: (i) CP (i.e., elements denoted by the green area in
Figure 3) and CIPSS (i.e., elements denoted by the red area in
Figure 3) parts in the model for the traditional slab system installation process and (ii) CP (i.e., elements denoted by the green area in
Figure 4) and HPCSS (i.e., elements denoted by the blue area in
Figure 4) parts in the model for the targeted slab system installation process. In both figures, (1) common work activities (depicted as CP1 to CP10 in
Figure 2) are represented by Nodes 1 to 17 in
Figure 3 and
Figure 4, and (2) indigenous work activities (explained based on CIP 1 to 9 and HPC 1 to 6 in
Figure 2) are represented by (i) Nodes 18 to 30 in
Figure 3 for CIPSS and (ii) Nodes 18 to 27 in
Figure 4 for HPCSS. Moreover, it is assumed that construction for a floor is completed after the completion of concrete curing and form removal tasks, which follow tasks of concrete placing work. However, both simulation models did not include two tasks after concrete placement, namely, concrete curing and form removal works, because it is not necessary to input specific resources for the concrete curing work, and this otherwise creates noise in analyzing other labor crew productivities and idleness. In addition, form removal work cannot be initiated for a long time (i.e., typically 4 to 13 working days), until concrete curing is finished, and this subsequently influences crew productivity measurement.
4.2. Simulation Model Implementation and Validation
Here, we remark that it is necessary to define information related to duration and resources with CYCLONE modeling elements COMBI, NORMAL, and QUEUE nodes to implement the CYCLONE simulation. The coding directions of CYCLONE indicated that (i) the duration data for each work task that is connected to another, in conjunction with the precedence relationship, is located at the COMBI or NORMAL nodes and (ii) the resource type and quantity of each work task are located at the QUEUE node. Information related to duration and resources is stochastically collected based on construction records or specifications of each task while the duration data of newly adopted work tasks are often gathered based on expert opinions (i.e., a deterministic method). Extant research indicates that it is possible to acquire reliable data on work duration if it is possible to derive various probabilistic distributions including normal, beta, and triangular distributions from raw data. Furthermore, it is widely known that a beta distribution is appropriate for work duration simulation data [
18]. However, it is not possible to follow the beta distribution in this study due to practical limitations in terms of the low number of HPCSS cases. Conversely, a triangular distribution is not significantly affected by the number of samples in the data, and thus, it can ensure that the collected data is reliable and accurate [
17,
19,
20].
Therefore, duration data of work tasks on CP and CIPSS are set by using a triangular distribution based on information to construct the 2nd, 3rd, and 4th floors of the building of interest (“case building”). Similarly, duration data of the work tasks on HPCSS are defined using information on installing 16 half-PC slabs on the 5th floor. In addition, resource data for each work task corresponding to CP, CIPSS, and HPCSS are derived from construction records of the case building.
Table 2 lists the resource and duration input data for each work task shown in
Figure 3 and
Figure 4. For example, from the table, we note that marking (Node 2) and horizontal stand installation (Node 4) most probably required 8 h based on the construction record of the case project. In addition, the duration of concrete placement (Node 27 in HPCSS and Node 30 in CIPSS) is set at 8 h by comparatively using a deterministic method.
The simulation was implemented with the simulation models, and input data represented by
Figure 3 and
Figure 4 and
Table 2 to analyze the construction productivity for types of work tasks and resources. The simulation results show that (i) first, the required cycle time for constructing each floor is calculated as 174.6 h and 103.3 h for CIPSS and HPCSS, respectively, (ii) construction productivity corresponds to 0.0057 (cycle/h) and 0.0097 (cycle/h) for CIPSS and HPCSS, respectively, and (iii) finally, these results can be interpreted as the delivery of higher productivity by HPCSS. A detailed explanation of these results is described in the “Findings and discussion” section.
In order to conduct a simulation study of the effects of HPCSS on construction productivity, it is necessary to verify if the developed simulation model can adequately reflect actual construction data [
20]. In the study, this verification was conducted based on comparing two types of data, namely, “collected data” from the actual case and “simulated data” from each model. That is, the study explores the following: (i) the extent to which the actual and simulated durations are identical for each work task and (ii) the manner in which events during the simulation chronologically occurred when compared with the actual construction process.
Table 3 lists the percentage difference between the simulated and actual durations for each work task. For example, with respect to Node 2 (i.e., “marking” in
Table 2), the simulated and actual durations correspond to 7.9 h and 8.0 h, respectively, and subsequently, the percentage difference of sum is estimated as 1.935% (i.e., ┃0.1019 − 0.1000┃/0.1000 × 100). A similar method is used to determine that after calculating the percentage difference of the sub-totals for all tasks (i.e., Nodes 2 to 30 for CIPSS and Nodes 2 to 27 for HPCSS), the lowest value corresponds to 0.443% for Node 27 for CIPSS, while the highest value corresponds to 23.871% for Node 6 for CP. Furthermore, (i) Nodes in the CP process (i.e., Nodes 2 to 17) exhibit an average difference of 85.723%, (ii) Nodes in CIPSS display an average difference of 3.096%, and (iii) Nodes in HPCSS exhibit an average difference of 12.710%. The results indicate that the developed models could be interpreted as being reliable in terms of the accuracy of construction duration.
Figure 5 compares the simulation results and actual case records, which enables us to examine as to whether the operation of the developed simulation model is identical to the actual work process. That is, the figure lists the HPCSS events that (i) are chronologically completed during the implementation of simulation (i.e., the top panel of the figure) and (ii) chronologically reported from actual construction records (i.e., bottom panel of the figure). In addition, based on COMBI (Nodes 19, 22, and 27) and NORMAL (Nodes 20, 23, and 24) elements with defined durations, the figure captures the simulated events from the initial task of half-PC installation (i.e., Node 19 in
Figure 4) assuming that the initiation time of node 19 is converted to zero. The principal results of comparison are as follows:
As shown in the figure, the installation of the first HPC slab unit is performed in conjunction with the following five tasks: HPC slab lifting (chronological list 1), HPC slab installation (chronological list 2), rebar delivery (chronological list 3), beam rebar installation (chronological list 5), and slab rebar installation (chronological list 7). Under this condition, the simulation time for the aforementioned tasks corresponds to 3.6 h while actual construction time for the first HPC slab unit from the record corresponds to 4.0 h, as denoted by “A” in the figure. Similarly, the installation work for the second HPC slab unit is completed in 6.6 h as per the simulation result and 6.0 h as per the actual record (Refer to “B” in
Figure 5). This result suggests that the time to complete the first slab as measured by the simulation is less than the time indicated by the actual record (i.e., 0.6 h), while the time to complete the second slab unit in the actual record is less than that in the simulation by a maximum of 0.6 h. In addition, the time to complete the installation of the last slab (i.e., 16th slab) is 41.2 h as per the simulation and 48 h per by the actual record. According to Hong et al. [
19], a simulation model developed using the actual data can yield (i) productivity rates that are closer to that in an ideal situation and (ii) lower uncertainty. Subsequently, the results of the simulation are superior in consistency when compared with those obtained from the actual data. From this viewpoint, the actual record reveals a buffer between the two works, as indicated by “C” in
Figure 5, although this buffer does not exist in the simulation result. Consequently, this indicates that the simulated completion time of the final unit is less than that in the actual record.
Moreover, the simulation result indicates that the work tasks for the first and second units of slabs are simultaneously ongoing, and this is also observed in the operations from the actual construction record (Refer to “D” in
Figure 5). Thus, the aforementioned results indicate that the developed simulation models are sufficiently accurate to be of value in further analysis.