On-site factories to support lean principles and industrialized construction

The main advantage of prefabrication is that it allows production and assembly of materials to take place in an environment that is usually not influenced by any external factors that generally prevail on construction sites, such as weather or dirt. Further, machinery used in prefabrication are highly specialized, and its specifications are tailored to the production requirements for producing output with specific characteristics and features. The condition of stationary, temporary, unique fabrication, as construction is generally defined, is no longer valid. Instead, prefabrication allows for leveraging economies of scale, reducing material and production costs by scaling and increasing production output while reducing construction time (Neelamkavil 2009).

Construction site management becomes less stressful and more efficient when some of the tasks are performed in the prefabrication phase, and fewer activities have to be done on the construction site itself (Väha et al. 2013; Rauch et al. 2015). One of the benefits is that common climate factors, which can have a detrimental impact on the quality of non-weather resistant materials such as timber, do not influence production schedule and quality (Tam et al. 2007; Richard 2005).

Improved environmental performance can be achieved by reducing construction waste caused due to poor site management practices (e.g., poor handling of materials or inadequate protection of finished work), lack of environmental awareness (e.g., less-standardized designs), damage during the delivery of products (e.g., method of transport), or rework due to poor quality (Tam et al. 2005). One example of the improved performances is the effective increase in material consumption from spontaneous solutions regarding design adaptation on structural panels. The cut off material ends up as waste that cannot be recycled for other production tasks and is therefore just thrown away. It is produced, transported, and stored and ends up being burned.

Statistically, the construction industry is one of the most hazardous industries in many countries (Wang et al. 2006; Forbes 2013). But, through prefabrication, many activities that are typically carried out on-site are undertaken in a safer and more stable working environment (e.g., on even, solidified or sealed factory grounds). This significantly reduces the likelihood of accidents and waste that occur due to external factors or adverse working conditions (N.P.C.A.A. 2017).

Even though prefabrication techniques have great potential to improve the sustainability of construction projects, they also have some disadvantages (Mokthar and Mahmood 2008). One of the main disadvantages is the inflexibility to amend orders due to the nature of the pre-fabrication concept and the fact that production plants are only seldom located near construction sites (Staib et al. 2008). Generally, the long distances between a factory and the construction site result in high order lead times since the ordered product has to be manufactured and transported to the construction site. The logistics of transporting prefabricated materials can be challenging because an efficient route to the construction site must be determined. Order modifications that arise in the short term lead to even more extensive time delays and increase costs for both supplier and client (Popovic and Winroth 2016; Tam et al. 2007; Panjehpour et al. 2013; Rauch et al. 2015). In addition, local jurisdiction can limit allowable routes for the transportation of oversized and often hazardous loads.

The costs induced by long transportation times between prefabrication plants and construction sites often have a tremendous economic impact on the overall product cost and result in higher risk related to damages. Furthermore, some remote areas may not even have reliable infrastructure and transport capacities, which may make it more difficult and costly to transport materials to their designated destinations (Weckenmann, n.d.).

The higher the costs for operators involved, the greater the needs for the production volume to maintain the same level of feasibility. This means that, due to the geographical separation of a specific construction process, such as the assembly of wall panels, additional operators are required to execute the same process. In addition to these disadvantages, off-site fabrication will have a great toll on local labor.

The advantages and disadvantages for off-site prefabrication stated in sub-section 2.2.1 show that it has already led to leaner construction. The disadvantages can be considered as possible opportunities for potential improvements. The following list introduces these opportunities and explains why they can be faced easily when establishing on-site prefabrication.

A work breakdown structure (WBS) of the SIP manufacturing process was defined through a video analysis showing a typical SIP manufacturing process.

Time studies, reference projects, and expert opinions are used to determine task times. These may vary as they depend on the level of automation at which they are executed, and other factors such as labor participation, motivation, and physical abilities (Erlach 2013). Therefore, each task was considered to have a range of possible durations, which was modeled using a beta-distribution with α = β = 4. Manual tasks that could be recorded in the video analysis were considered with minimum and maximum durations within a range of 15% and the tasks that were at least partly automated were considered to have a constant duration. Tasks that were not undertaken in the process procedure in the video analysis and whose durations had to be assumed based on expert opinion were modeled with higher duration ranges of 25% in both directions, maximum and minimum (Table 1).

The next step entails defining the relationships between the tasks. These “logical relationships” can either be arranged in a parallel or sequenced manner and are strongly dependent on the number and quality of resources used. The general logical layout for a MgO SIP standard panel consists of different tasks. There were 14 tasks in total for the process used in this study, and are summarized in Table 1.

Takt time is defined as the ratio of total available production time to the time required to produce one unit (or customer demand). In other words, it is the rate at which a finished product needs to be completed in order to meet the customer demand. This element is of great importance because the assembly line needs to be adjusted to reach a given takt while maintaining line efficiency (Erlach 2013). Since the takt time depends not only on work efficiency but also on the number of resources used, a range of takt times is taken into account in order to determine a reasonableALD.

The takt time and the demand rate are highly correlated. As customer demand changes, the takt time also changes, and it requires a rebalancing of the production line (Chang et al. 2013). It is used to ensure that all the work to be performed for a given product in a given station is completed in the specified time before proceeding to the next or subsequent station (Qattawi and Chalil Madathil 2019).

During the flow of material through the assembly line, the material is processed at workstations that may include resources as machines or operators. The number of workstations required to process a product strongly depends on the time it takes to manufacture one piece and the delivery time of the products as demanded by the customer, i.e., the takt time. The number of operators required at a specific workstation is dependent on the level of automation applied to the process. The choice of the number of workstations may affect both the cycle time and the line efficiency since the evenness of task distribution is limited most of the time by the number of workstations among which the tasks are distributed.

Line balancing affects how well material flows through a production line when aiming for uniform workstation cycle times that are close to customer takt time and minimizing the number of workstations and operators or, in sum, maximizing economic efficiency (Thomopoulos 2014; Erlach 2013). Two different measures can be used to adjust line balancing:

Shifting of work tasks from one workstation to another

The shifting of work tasks from one workstation to another can also be viewed as assigning more than one working task to one workstation. This measure is quite simple and has a significant effect on balancing. Stacking together work tasks at different workstations may also lead to a reduction in the number of workstations, which reduces not only the cost but also the required area for the production line. The second measure can be achieved with a higher level of automation and, in some cases, an increasing number of operators. Higher levels of automation may lead to tasks being finished very quickly or becoming obsolete. In summary, the first one consists of shifting of work tasks from one workstation to another and the second one consists of upgrading/downgrading of workstation resources in terms of the level of automation to increase/reduce the workstation operation time.

Generally, the on-site factory can be located temporarily in rented space near the construction site or at the construction site itself through mobile factory structures (e.g., shipping container) or on the growing building structure (e.g., climbing concrete factories) (Rauch et al. 2015). The shell has a significant effect in terms of its mobility, which is referred to as “outer mobility.” Outer mobility is, of course, strongly related to “inner mobility,” which refers to the mobility of workstations, machines, and general interior installations related to a factory. While providing a certain level of inner and outer mobility, a shell design has to meet requirements to ensure functional reliability and work safety (Erlach 2013). The shell structure has to satisfy static and climatic requirements. Depending on the shell structure system, the shell has to be transportable or easy to assemble. If it is transportable, it must follow road traffic laws and regulations. If the shell is a system that needs to be assembled, it has to be suitable for multiple assembly and disassembly without wear and tear, or otherwise would lead to a less reliable system. In addition, the shell structure has to be capable of withstanding stress from rain, snow, ice, and wind to be workable for the long term and to provide functionality regardless of climatic conditions. Moreover, the shell needs to be soundproof up to a certain point (Wiendahl et al. 2014).

The relationships between the tasks and their arrangement (e.g., in parallel or sequenced manner) were determined to define the process logic. Figure 3 shows the process logic containing the main tasks for the manufacturing of standard MgO SIPs.

Fig. 3

It consists of 14 working tasks (Table 1). The “add timber frame” task (ID 14 in Table 1) can only start when the other three processes (i.e., prepare timber framing, produce MgO panel (1), and produce MgO panel (2)) are finished. The “produce MgO panel” process is executed twice in this layout and consists of the tasks shown in Figure 4.

Fig. 4

Task times were determined from a video analysis showing a typical SIP manufacturing procedure and crosschecked with experts’ opinions and previous studies. The task times vary depending on the level of automation. Other factors include labor participation, motivation, and physical abilities (Erlach 2013). Depending on the scope and the certainty with which task times can be stated, either stochastic or deterministic approaches were considered for approximation. However, for short discrete construction productions, the duration variances tend to be small, and there are not numerous varieties for their distributions. The tasks and their corresponding minimum and maximum durations for this study are shown in Table 1. Those durations were used in the modeling template of Simphony.NET (Simphony.NET 4.6, release Build 4.6.0.272 2017-08-11). Tasks that are considered to have a range of possible durations were modeled using a beta-distribution. A beta-distribution was used as it generally provides a better fit for activity durations (Shtub et al. 2004). For simplicity, α and β were set to 4 to get a symmetrical continuous distribution and facilitate the simulation. The lower and upper bounds of the frequency distribution for the different tasks were set to t_emin and t_emax, respectively. Manual tasks that were included in the video analysis (Task IDs 1 through 6 in Table 1) were considered with minimum and maximum durations within a range of ±15% in order to account for variations originating from the motivation and skill level of the workers. Tasks that were at least partly automated were considered constant in their duration (Task IDs 7 through 12 in Table 1). For the tasks that were not part of the video analysis (e.g., Task IDs 13 and 14 in Table 1), durations were estimated based on expert opinion and modeled with higher ranges (±25%) to account for extra uncertainty.

Determining the assembly line design in terms of distributing working tasks to workstations depends on the sum of the working task times, which was found to be 360 s (6 min) on average (see Table 1). The takt time from a conventional operation is the summation of all the times for the necessary tasks. In this case, the time it takes to hook a finished panel to the crane, place it, install all required elements, unhook it from the crane, and bring the crane back into the starting position in order to hook the next panel was 11 minutes. Figure 5 shows the relationship between the number of required workstations and takt time. A maximum of 1 min and a minimum of 10 min were considered for the takt time, since the assembly line has to consist of at least one workstation, and takt times—with reference to the takt time for one customer at the time of being supplied—lower than 1 min are highly unlikely. Such takt times can only be achieved if more than ten customers are supplied with SIP panels at the same time. Here, customers are defined as separated working processes that run in parallel and require SIPs. Such situations can occur if a construction site uses more than one crane and labor group for SIP installation, which could be the case of bigger projects that typically consists of more than one building and/or construction cranes.

Fig. 5

As shown in Figure 5, takt times greater than 6 min require an assembly line with only one workstation to process SIPs on time, referred as assembly line variant 1 (AL-1). In addition, takt times from 6 to 3 min from 3 to 2 min, and 2 to 1 min are considered for variants AL-2, AL-3, and AL-4, respectively (Table 2). Lower takt times may occur from higher productivities by the producer or by a higher number of customers demanding SIPs at the same time. Higher productivities mainly originate from above-average skilled labor or higher levels of automation (e.g., the use of nail guns, automatic screwdrivers, or jigsaws). The influence of the learning curve due to the frequent-changing of operators is a significant factor, but for purposes of this study, it was simplified by assuming that the learning curve effects were considered for the average values used.

Table 2 shows the different assembly line configurations and information related to the simulations done in order to determine the best fit regarding labor cost (i.e., number of workers), total weight and size (to ensure mobility of the assembly line), average cycle time, distribution of tasks to workstations (WS), required space for the setup, and total utilization.