Eficient facility layout and risk-aware planning are critical to both productivity and safety in manufacturing environments. This study examines the impact of workspace dimensions and performance on risk and production rates in a dry transformer production line. A nonlinear model is developed to estimate operators' failure rates due to stress caused by workspace dimensions and workload, which are dependent on facility layout. The model is evaluated using discrete-event simulation (DES) with Arena software and a case study. Scenario analyses examined variations in workspace dimensions, and the results indicated that increasing maneuvering area and optimizing spatial allocation can reduce operator failure rates while improving production output. Average weekly production increased from 142 units (baseline) to 154 units (an 8.4% increase), while operator uptime (F) rose from 16.92 to 28.21 hours, from 13.54 to 23.69 hours, and from 16.92 to 28.21 hours across the three main work areas on the production floor. These findings show that strategic layout adjustments enhance safety and productivity, enabling manufacturers to design risk-informed workspaces that reduce failures and improve eficiency.
Manufacturing enterprises often encounter limitations in both short-term and long-term planning,
which negatively impact their competitiveness and long-term sustainability [
In this context, discrete-event simulation (DES) is a strategic tool for analyzing scenarios and
supporting operational decisions [
Beyond eficiency gains, safety can also be enhanced through predictive risk assessment, as DES
enables the simulation of failures, workflow interruptions, and hazardous scenarios to address
vulnerabilities proactively [
The remainder of this manuscript is structured as follows: Section 2 introduces the proposed risk model; Section 3 presents the case study and data collection; Section 4 describes the DES model developed in Arena®; Section 5 reports validation and results with the risk model; and Section 6 provides conclusions and directions for future research.
All operations within manufacturing processes are fundamental to ensuring efectiveness; identifying
risk factors allows companies to maintain infrastructure, equipment, and other production elements
in optimal condition [
Proper facility layout is consistently identified as a key factor in reducing accident risk [
The present work advances this field by modelling risk as a function of the available space for human maneuvers and integrating this risk model with DES to estimate its impact on production and resource performance dynamically. This approach enables a more comprehensive evaluation of how spatial constraints and layout decisions influence both safety and operational eficiency on the manufacturing lfoor.
Accidents and failures in industrial systems result from the combined influence of multiple operational
drivers, each contributing to overall risk through distinct mechanisms. The number of workers during
interval () increases the frequency of potential interactions; as the workforce size grows, the
probability of interference, hand-ofs, and uncoordinated actions rises accordingly. The processing
pace during interval () determines exposure intensity, since higher throughput per worker elevates
cognitive load and fatigue, thereby increasing the likelihood of unsafe actions. Spatial constraints,
represented by the crowding factor during interval ( = 1/), limit maneuverability and elevate the
chance of collisions or restricted movements. Finally, resource utilization during interval ( ) captures
the extent to which the system operates near its adequate capacity. As utilization approaches unity,
operational bufers diminish, queues lengthen, and time pressure intensifies, producing a nonlinear
escalation in accident probability. Note that these mechanisms act multiplicatively rather than additively:
small, simultaneous increases in workforce size, pace, crowding, and utilization may combine to
generate a disproportionately large rise in risk. To capture this behavior, the model adopts a log–linear
specification that transforms into a multiplicative power-law form after exponentiation [
By defining 0 as the baseline accident rate under reference operating conditions (0, 0, 0) the log-linear mean accident rate during interval ( ) is represented as: ( ) = ( 0) + 1( 0 ) + 2( 0 ) + 3( 0 ) − (1 − ) (1)
This logarithmic formulation ofers statistical tractability, ensuring dimensional consistency and interpretability. The coeficients 1, 2, 3, and represent the proportional change in accident risk resulting from a proportional change in the associated factor (, , , and , respectively). Table 1 provides additional details on the dimensions and characteristics of these variables and terms.
Exponentiation of both sides of (1) leads to the multiplicative power-law form: Overview of variables and dimensions of the proposed risk model. within a single framework. Existing models do not simultaneously quantify the impact of facility layout decisions on both outcomes, and metrics such as efective workspace and worker density—factors directly influencing accidents and eficiency—are seldom incorporated. Furthermore, DES has rarely been extended to include space-dependent risk variables, and validation eforts using current empirical data remain limited. In this context, the model presented in (1) and (2) represents a significant contribution by bridging these gaps and providing a unified approach to assessing safety and eficiency. (2)
The company in the case study specializes in the manufacture of dry transformers. The analysis focuses on the materials warehouse area, where the company is proposing the development of a production line for the final assembly of transformer units (coils). Key operations in this line include unpacking, subassembly, part joining, quality inspections, product identification, electrical testing, insulation application, and final packaging. The process flow is illustrated in Figure 1.
For each operation, ten samples were collected to support the statistical modelling process. To account for variability in activity times, a 15% fatigue factor was incorporated. Table 2 presents the iftted probability distributions associated with the time required to complete all activities.
As production rate reference, four units of Part A and Part B are received each hour for processing within an eight-hours shift. This leads to a weekly input of 4 units * 8 hours * 5 days = 160 units approximately. However, the actual production rate is estimated to be 28 units per day, which translates to approximately 140 units per week.
Rockwell’s Arena® software was used for modelling and analyzing the case study. The software includes modules for diferent processes and operations, animation, execution, and results visualization. The general simulation model is presented in Figure 2. In accordance with Figure 1, the process begins with the arrival of Part A and Part B components. Both components arrive at a rate of four parts per hour during the 07:00 to 15:00 shift. Figure 3 shows the configurations of the "Create" modules. Meanwhile, Figure 4 illustrates the configurations of the “Process” models, considering the operators and processing times outlined in Table 2.
As presented in Figure 2, "Decide" modules “Acceptable 1” and “Reworkable” are added to verify whether the product passes the “Quality Inspection” process. Statistics report that 90% of the units pass the “Quality Inspection” process while 10% are sent to “Rework”. Of the reworked units, 10% are discarded. The “Decide” modules “Acceptable 2” and “Successful Lab Tests” are added to verify if the product complies with the “Hi Top TTR” test. If it complies (90% of cases), it advances to the following process. If it fails, it is sent to subsequent laboratory tests. If it passes the retest (in approximately 90% of cases), it proceeds to the following process; if not, it is discarded.
To support the analysis of the current system and guide improvement initiatives, the virtual model
underwent statistical validation. A total of 30 weeks of output data were collected from the real system
(1), and 30 corresponding replications of weekly production were generated using the simulation
model (2). Validation was performed using Welch’s t-test [
Table 3 presents the results of the validation with = 47 degrees of freedom and a significance level = 0.01 , yielding a critical value of /2, = 2.687. Since |0| ≤ /2, the null hypothesis cannot be rejected, indicating that both systems exhibit statistically similar behavior.
average output (1 = 142).
productivity. Estimation of (uptime) for the virtual system with the risk model ( 1 = 0.8, 2 = 0.6, 3 = 1.0, = 2) in A: Zone 1 from 12 2 to 20 2, Zone 2 from 50 2 to 70 2, and Zone 3 from 15 2 to 25 2. As presented, an increase in can be achieved by expanding the area available for human maneuvers.
With the updated risk models (Table 5), the virtual model yields 154 units on average, an 8.4% increase over the real system, suggesting that investment in parameter A could reduce failure rates and enhance Estimation of (uptime) for the virtual system with the risk model and increased ( 1 = 0.8, 2 = 0.6, 3 = 1.0, 0 Zone 1: 12 2 × 0.20 = 2.4 2
Zone 1: 20 2 × 0.20 = 4.0 2
The results of the simulation model validation indicate that the virtual model is more representative of the real system, as it does not exhibit statistically significant diferences compared to the case study data. The inclusion of probabilistic distributions for processing times is confirmed, and the consideration of failures allows for capturing the variability and interruptions of the production system more accurately. The data show that the failure rate ( ) decreases considerably when the area available for human maneuvers () increases, translating into improvements in the average production rate (e.g., increased average weekly production from 142 to 154 units). These results demonstrate the importance of considering physical limitations in simulating production processes.
Also, this work demonstrates that DES is a practical approach for analyzing and optimizing manufacturing systems, providing a virtual environment to evaluate diferent operational scenarios and their impact on both productivity and safety. By modeling facility layout and incorporating spatial risk assessment, the research establishes that the area available for human maneuvers is a critical factor in reducing operational failures and improving overall system reliability.
Beyond the dry-transformer case study, the -model and DES workflow are scalable. Because the model variables—workforce size, work pace, available space, and utilization—are general descriptors of human–space interaction, they can be recalibrated for multi-line facilities or adapted to non-assembly domains. Likewise, the modular design of DES software, such as Arena, allows the framework to be extended to larger systems.
Future work will advance in three directions: (a) sensitivity analysis of the power-law formulation, (b) integration of real-time data from industrial sensors and IoT devices to enable dynamic risk assessment and adaptive process optimization, and (c) exploration of parallel or cloud-based simulation to further enhance scalability for enterprise-level applications. These extensions will strengthen the model’s applicability and support continuous monitoring, predictive maintenance, and rapid response to operational disruptions.
The authors have not employed any Generative AI tools.