Enabling Manual Workplace Optimization Based on Cycle Time and Musculoskeletal Risk Parameters

Abstract

Recently the concept of Industry 5.0 has been introduced, reinforcing the human-centric perspective for future industry. The human-centric scientific discipline and profession ergonomics is applied in industry to find solutions that are optimized in regard to both human well-being and overall system performance. It is found, however, that most production development and preparation work carried out in industry tends to address one of these two domains at a time, in a sequential process, typically making optimization slow and complicated. The aim of this paper is to suggest, demonstrate, and evaluate a concept that makes it possible to optimize aspects of human well-being and overall system performance in an efficient and easy parallel process. The concept enables production planning and balancing of human work in terms of two parameters: assembly time as a parameter of productivity (system performance), and risk of musculoskeletal disorders as a parameter of human well-being. A software demonstrator was developed, and results from thirteen test subjects were compared with the traditional sequential way of working. The findings show that the suggested relatively unique parallel approach has a positive impact on the expected musculoskeletal risk and does not necessarily negatively affect productivity, in terms of cycle time and time balance between assembly stations. The time to perform the more complex two-parameter optimization in parallel was shorter than the time in the sequential process. The majority of the subjects stated that they preferred the parallel way of working compared to the traditional serial way of working.

1. Introduction

To stay competitive on the market, companies need to manufacture products with high quality and productivity. Simultaneously, companies need to offer production staff a sustainable work life. There has been shown a strong correlation between poor ergonomics and both productivity losses and quality deficiencies, e.g., in the form of assembly errors [1]. Hence, poor ergonomics in the production lines can result in major costs for companies. In the research of Falck et al. [1] the term ergonomics represents the human well-being perspective of the ergonomics definition. Ergonomics, or human factors, is defined as the scientific discipline concerned with the understanding of interactions among humans and other elements of a system, and the profession that applies theory, principles, data, and methods to design in order to optimize human well-being and overall system performance [2]. It is found that most research papers tend to address one of the domains at a time, i.e., either human well-being or overall system performance [3]. Despite the above-mentioned relationship between human well-being, productivity, and quality, the domains are commonly treated separately in industry when conducting factory, line, and workstation planning. In manufacturing and production planning, traditional line balancing is conducted with a focus on achieving a certain production volume capacity with a small amount of waste, e.g., walking. Most commonly, the parameter that is optimized is time. Since time and productivity are closely related, productivity (system performance) is the focus in traditional line balancing. Human well-being is considered in traditional line balancing but is rarely prioritized or explicitly denoted in the balancing activity. Instead, worker well-being is commonly in focus at later stages of the production development process. However, at these stages, constraints from the earlier optimization focusing on productivity means that there are limited opportunities for improvements and optimization of any parameters other than time. Such a serial process is not preferable, since it makes optimization of several parameters slow and complicated; hence, it has a negative effect on development time and is likely to result in suboptimal solutions. Industry typically focuses on concurrent and integrated engineering processes, where design and development, production planning, and related activities are carried out more or less in parallel [4]. The approach to performing activities in a parallel manner is also desired and beneficial in subtasks in the production planning process. Therefore, time balancing and worker well-being should be considered together in a parallel process to achieve an efficient production planning process. A parallel process is more likely to facilitate finding optimal solutions for both time and well-being; whereas a serial process is more likely to find solutions that are local optimums, i.e., suboptimal solutions that lack the consideration of the overall perspective.

Simulation and automation are elements in the Industry 4.0 transformation that effect manufacturing industries’ production systems and ways of working [5]. Recently the concept of Industry 5.0 has been introduced, reinforcing the human-centric perspective for future industry, complementing Industry 4.0 that focuses on automation and digitalization transformation [6]. Industry uses different kinds of software in the production development process to represent, analyze, and simulate production in order to create insights into productivity and plan production prior to implementation [7]. Accordingly, production/process engineers use software tools for factory and line balancing, in turn to optimize productivity. Some of these are specialized software tools, such as AVIX^TM, TiCon Takt^TM and Casat^TM, whereas others are parts of manufacturing software suites, such as Siemens PLM^TM and Dassault Systems^TM. In order to simulate digital manikins performing work tasks and, based on the simulation, evaluate aspects of well-being, industry can use digital human modelling (DHM) software [8]. Examples of DHM software tools are as follows: 3D SSPP [9], Siemens Jack [10], RAMSIS [11], and IPS IMMA [12]. Different DHM tools have different focuses and key application areas, and hence strengths and weaknesses depending on what they are to be used for. There are a number of review papers summarizing DHM software [13,14]. It is typically human factors engineers that perform simulations of human work and evaluate if well-being criteria are fulfilled. This means that, regarding use of simulation software, the consideration of productivity (time balancing) and human well-being are separated in the development process, involving different professions, and following a sequential process. Such a way of working easily results in a slow development process. Also, final solutions risk being suboptimal, since the manual optimization efforts to find successful solutions from a combined productivity and human well-being perspective become so complicated and intertwined. To optimize aspects of both productivity and human well-being in a parallel process is, however, not unique. The literature includes several publications where the optimization problem has been solved by the use of mathematical algorithms. For instance, Otto and Scholl [15] described how musculoskeletal risk assessment measures, e.g., NIOSH Lifting equation [16] and OCRA [17], were incorporated into an automatic line balancing problem. Akyol and Baykasoğlu [18] also used the OCRA method to minimize cycle time and musculoskeletal risks in a constructive rule-based random search algorithm. Such mathematical models and software are, however, rarely used in industry [19]. A major reason is that the line balancing problems in industrial settings are more complex than cases used in the literature mentioned. Hence, industry uses support software and manual balancing approaches. A few attempts to introduce manual human well-being and cycle time balancing tools have been made. ErgoSAM [20] was one of the first attempts to develop such a method and software. The software was used in industry for a while but disappeared. It has not been documented why, but reasons may be related to the usability of the tool and how it was integrated in the way of working at the companies. Perez and Neumann [21] reported that the usability of human factors software tools is of huge concern in order for the tools to be used by engineers and successfully be integrated into companies’ work processes. Also, ErgoSAM requires that MTM-SAM (sequence-based activity and method analysis) is used as the method for time estimation [22], which is of limited use in industry, especially in a global perspective.

In summary, current production balancing practices include efficient tools and methods for both cycle-time optimization and reducing musculoskeletal risk. However, these resources are not currently combined, resulting in a sequential work process. To make factory, line, and workstation planning more efficient, to ensure appropriate conditions for production staff, and to improve productivity and quality, the aim of this article is to suggest, demonstrate, and evaluate a concept that enables line balancing that considers two parameters in parallel: assembly time as a parameter of productivity (system performance), and risk of musculoskeletal disorders as a parameter of human well-being. The hypothesis is that this parallel method can deliver results comparable to the existing approach in terms of cycle time, while reducing musculoskeletal risk and accelerating the production preparation process. The experimental findings support this hypothesis.

2. Materials and Methods

2.1. Software Demonstrator

AVIX software version 5.0 (Gothenburg, Sweden) was used to demonstrate the concept. AVIX is a software that supports analysis of manual assembly processes by combining video analysis with time and motion studies based on clocked times or standard times such as MTM [23], MTM-SAM [22], and MTM-UAS [24]. AVIX software also supports import and export of data with DHM tools [25]. The applications of AVIX are in the fields of productivity, production development, production optimization, and production efficiency, which are all closely related to work measurement and time study.

A demonstrator version of AVIX was created, in addition to the regular functionality of the software, to facilitate the consideration of musculoskeletal risks based on static posture assessments using the rapid entire body assessment (REBA) method [26]. One aspect of optimizing an assembly line is to even out the workload between operators through transferring assembly activities from one operator to another, i.e., from one work position to another. Provided that the time estimation is conducted per assembly activity, the time-related impact of transferring an assembly activity between work positions is seen instantly in the balance chart. On the other hand, human well-being assessments are in general conducted per work position, not per assembly activity. Typically, posture and force assessments are conducted by an ergonomist observing the work performed at the work position and assigning the work position a score based on the characteristics of the work activities at the position. This is also typically the case when using the REBA method. The circumstances mentioned above make it hard to optimize an assembly line from a well-being point of view since the assessment of a work position becomes invalid when transferring assembly activities between work positions. Thus, in order to facilitate the balancing of time and well-being based on the same balancing element, i.e., assembly activity, the demonstrator is realized using a modification of the REBA methodology. The modification means that REBA assessments are conducted per individual assembly activity, not per work position as in the original REBA method. The REBA assessment scores per assembly activity are then aggregated using a time-weighted average REBA score to represent the well-being conditions of the work position.

To facilitate the balancing of assembly activities between different work positions in regard to both time and musculoskeletal risk, the demonstrator version of AVIX includes a graphical presentation that shows the total assembly time and the aggregated ergonomic score per work position, as well as the time and musculoskeletal risk score per assembly activity (Figure 1). Based on the graphical presentation, the user can distribute the assembly activities between work positions to achieve solutions that satisfy the production capacity requirements and, at the same time, distribute musculoskeletal load reasonably equally between work positions and, hence, between operators. The interface is interactive, and the user can make drag-and-drop operations directly in the graph (Figure 1).

2.2. Test Subjects

Thirteen test subjects were recruited. The requirement for participating in the test was prior knowledge of AVIX, balancing, and ergonomics, e.g., musculoskeletal risks. Figure 2 shows the number of years of experience in each field. All participants had deep knowledge of vehicle manufacturing and represented manufacturers located in Asia, Europe, and South America. Two test subjects did not have any formal training in balancing and four test subjects had no formal training in ergonomics. Six subjects had basic training in balancing, and seven in ergonomics. Two test subjects had formal training in balancing from university, and one in ergonomics from university. Three subjects had other formal training in balancing, and one in ergonomics. On a five-grade scale, with anchor points ‘basic’ and ‘expert’, the subjects’ average self-ranked experience in AVIX was 3.6, in the balancing domain 3.9, and in ergonomics 2.3.

2.3. Use Case

A use case was created with the objective that it should be generic but still include some complexities in regard to the balancing of manual assembly from both a time and a musculoskeletal risk perspective. The use case consisted of the assembly of a pedal car, as a simplified but valid representation of manual assembly work. Simplified products, such as, for instance, pedal cars, are frequently used to teach lean and production philosophy concepts [27]. The use case was described extensively with detailed work task instructions. In total, the use case comprised 78 assembly activities, referred to as tasks, most of which needed to be carried out in a specific sequence. A precedence graph illustrates the relation between the tasks (Figure 3).

The tasks with the same color close to each other need to be performed together, otherwise quality might be jeopardized (Figure 3a). By joining associated tasks, the total work can be grouped into 28 task groups, forming a simplified precedence graph (Figure 3b). Other than these constraints, the assembly sequence may be altered. These task groups were made available in AVIX, where all tasks had descriptions of what the worker should do, how to perform the task, and an explanation of why the task is performed. Each task, and consequently all task groups, also have a given pre-determined time, based on MTM/SAM. Furthermore, all individual tasks also have a given musculoskeletal risk score, based on a time-weighted REBA score. The assembly tasks and task groups that were considered simple, i.e., without complex body motions during assembly, were assessed manually using the REBA assessment tool in AVIX software (Figure 4a). The more complex tasks were simulated and assessed by REBA analysis using the DHM tool IPS IMMA (Figure 4b), and the REBA scores were then sent to AVIX using the import/export functionality that AVIX provides.

2.4. Procedure

The participants were invited to a short introduction meeting describing the experimental test, where it was clarified that their objective was to optimize the pedal car assembly line, consisting of three assembly stations, in terms of minimizing cycle time and maximizing worker well-being, i.e., minimizing musculoskeletal risk. The participants were asked to perform the optimization mission twice, using two different versions of the AVIX software. First, in Experiment 1 they used the standard version in which time is noted (in seconds) for each task and musculoskeletal risks are considered indirectly and with expert support. Second, in Experiment 2 they used the AVIX demonstrator version where both time and musculoskeletal risk score are noted for each task. Both AVIX versions contained the prepared pedal car assembly use case, with an assembly line involving three work positions, here referred to as stations; all task groups were placed in station one. It was then up to each participant to drag and drop the tasks and task groups from one station to another, taking into account the precedence relations and the overall objective. The participants were given deadlines for performing the experiments. These deadlines were aimed at allowing the participants to work concurrently with their normal workload, and to ensure results could be received in a timely manner.

For the first optimization in Experiment 1, the participants were aiming for the best possible balance considering assembly time and were provided with the AVIX file without musculoskeletal risk data visible. The musculoskeletal risk considerations were made indirectly based on their own experience. The completed balancing optimization mission was handed in, and the musculoskeletal risk analysis was conducted by an ergonomics expert. The users were provided with some feedback from the ergonomics expert. The feedback pointed out which station performed worst in terms of musculoskeletal risks. The participants were then instructed to try to include this feedback for the completion of the balancing mission, still using the standard version of AVIX. When Experiment 1 was finished and reported, the participants instead used the demonstrator version of AVIX in Experiment 2, the one containing musculoskeletal risk scores per assembly task and functionality, to indicate musculoskeletal load on the station level. Using the demonstrator version, the participants could now see both the time and musculoskeletal risk-related effects of moving tasks or task groups between stations. Hence, they could easily keep track of each station’s time-weighted average REBA score. When the second balancing mission was finished and reported, the participants were asked to fill out a questionnaire. The questionnaire included questions about the set-up and the way of working in Experiment 1 and Experiment 2. The participants were also asked to compare the two ways of working.

2.5. Data Analysis

The balancing results from the first and second reports from Experiment 1 and the report from Experiment 2 were analyzed. The key performance indicators (KPIs) calculated were the system performance measured cycle time and the musculoskeletal risk measures REBA Sum and REBA Max. REBA Sum and REBA Max have different objectives: to reduce musculoskeletal risk across the total assembly line and reduce the peak ergonomic risks in the assembly line, respectively. The Wilcoxon signed-rank test was used to test if a significant difference (p < 0.05) exists between: (a) the first report from Experiment 1 (denoted Test 1:1), where subjects minimized cycle time and considered musculoskeletal risks indirectly based on their own experience; (b) the second report from Experiment 1 (denoted Test 1:2), where subjects optimized cycle time and musculoskeletal risks after having received feedback from an ergonomics expert; and (c) the report from Experiment 2 (denoted Test 2:1), where participants optimized cycle time and musculoskeletal risks in parallel with support of the AVIX demonstrator version.

3. Results

The experiment comprised three test parts: 1:1, 1:2, and 2:1. The times it took for the participants to carry out the optimization missions are given in

Table 1. Optimization time, cycle time, and time-weighted REBA scores (Sum and Max), per station, for Tests 1:1, 1:2, and 2:1. Differences between the tests are presented. Significant differences on p < 0.05 level is indicated by ***.

				Time Weighted REBA Score
Test		Optimization Time [min]	Cycle Time [s]	Sum	Max
A1	Min	10	113	8.9	3.2
	Mean	44	117	9.0	4.1
	SD	34	5	0.1	0.5
	Max	120	127	9.3	4.6
A2	Min	5	113	9.0	3.2
	Mean	22	118	9.1	3.8
	SD	14	4	0.1	0.5
	Max	60	124	9.5	4.8
B1	Min	5	114	8.6	3.0
	Mean	33	128	8.9	3.4
	SD	20	22	0.2	0.3
	Max	75	175	9.1	4.1
A1–A2		22	−1	−0.1	0.3
A1–B1		11	−11 ***	0.1 ***	0.7 ***
A2–B1		−11	−10 ***	0.2	0.4 ***

. The time to perform the time balancing considering musculoskeletal risks indirectly (Test 1:1) was on average 44 min. After feedback regarding musculoskeletal risks (Test 1:2), the second rebalancing required on average 22 min less. In total, the serial process to optimize time and musculoskeletal risks (Tests 1:1 and 1:2) took 67 min on average with a standard deviation of 41 min. In the parallel balancing process optimizing both time and musculoskeletal risks (Test 2:1), the average time was 33 min with a standard deviation of 20 min. The results for each test and each KPI, i.e., cycle time, REBA Sum, and REBA Max, are presented in Table 1. The shortest cycle time, 113 s, was achieved in Test 1:1. The lowest REBA Sum and REBA Max were achieved in Test 2:1, with scores of 8.6 and 3.0, respectively. Figure 5 and Figure 6 show KPI values for each of the thirteen participants’ balancing solutions from Tests 1:1, 1:2, and 2:1, where Figure 5 plots REBA Max versus cycle time and Figure 6 plots REBA Sum versus cycle time. The balancing solutions that gave the shortest cycle time and the lowest REBA Sum and REBA Max are located on the Pareto front in Figure 5 and Figure 6 and are possible optimal solutions. However, there are other solutions, not the extreme ones, on the Pareto that can be considered possible solutions.

Comparing the cycle time, REBA Sum, and REBA Max, there were no significant differences between the results from Tests 1:1 and 1:2 (Table 1). A significant difference was found in cycle time, which increased (p = 0.031), and REBA Max, which decreased (p = 0.003), when comparing the results from Tests 1:2 and 2:1 (Table 1). A significant difference in cycle time, REBA Sum, and REBA Max was found when comparing the results from Tests 1:1 and 2:1 (Table 1), where cycle time increased (p = 0.047), and both REBA Sum (p = 0.013) and REBA Max decreased (p < 0.001). In summary, all the results support the hypothesis that the parallel method can deliver results comparable to the existing sequential approach in terms of cycle time, while reducing musculoskeletal risk and accelerating the production preparation process.

In the post-experiment questionnaire, six of the thirteen subjects stated that Test 1:1, considering only cycle time was the normal way of working (

Table 2. Summary of comments after the tests, and comments after comparing the serial way of working with the parallel process.

	No of Comments Out of 13	Comments
After test 1:1 and 1:2	6	considering only cycle time is the normal way of working
	3	difficult to let go of usual reality restrictions
	3	hard to work with the precedence graph
	4	hard to work with uncommon tasks
	6	normally we only consider cycle time
After test 2:1	8	the way of working is interesting and useful
	4	difficult to optimize several objectives in parallel
	1	not have enough ergonomics knowledge
Comparing test 1:1 and 1:2 with test 2:1	9	preferred the parallel way of working (2.1)
	3	results were as expected: longer cycle time and improved human well-being in 2.1
	2	cycle time becomes too long when musculoskeletal risks are considered at the same time

). Three subjects mentioned that it was difficult to let go of usual reality restrictions. Three subjects mentioned that it was hard to work with the precedence graph. Four subjects mentioned that it was hard to work with uncommon tasks. The subjects also commented on the demonstrator version of AVIX used for Test 2:1 (Table 2). Eight of the thirteen subjects described the way of working in Test 2:1 as interesting and useful. Four stated that it was easy to work with. In contrast to these four subjects, three subjects stated that it was difficult to optimize several objectives in parallel. One subject admitted that he did not have enough knowledge of ergonomics. When comparing Experiment 1 (Tests 1:1 and 1:2) with Experiment 2 (Test 2:1), nine of the thirteen subjects stated that they preferred the parallel way of working (i.e., as in Experiment 2) (Table 2). Three subjects mentioned that it was an easier way to consider musculoskeletal risks compared to the traditional way of working with observation methods and expert evaluations. Three subjects reflected that the results were as expected: a longer cycle time and improved human well-being. Two subjects claimed that the cycle time becomes too long when musculoskeletal risks are considered at the same time.

Final Demonstrator

After the tests, a final demonstrator was developed based on the subjects’ responses in the questionnaire, as well as additional experiences from the experimental tests. In Figure 7, the black line represents the customer takt time, i.e., how often a customer orders a product on average, taking into account the available production time. The orange line in Figure 7 represents the maximum acceptable musculoskeletal load. The musculoskeletal risk assessment is presented in green, yellow, orange, and red depending on the score per assembly activity and aggregated per work position. Green is the lowest score, and red is the highest score. The portion of the aggregated musculoskeletal risk score per work position in the green range is shown below the customer takt time. The portion of the aggregated musculoskeletal risk score that is in yellow, orange, and red is indicated from the takt time and upwards. The interpretation of the figure is that, as long as the bar representing the musculoskeletal risk score is below the orange line, the work position is acceptable from a musculoskeletal risk perspective. The bars for both time and musculoskeletal risk score are updated when a task is transferred from one work position to another. This allows the user to optimize the production line considering both parameters. Note that the vertical position of the orange line is controlled by a setting in the software to allow configuration of the demonstrator to align with the chosen methodology and corresponding threshold. In addition, a specific human well-being method may prescribe other interpretations, for instance, that all activities classified as red must be eliminated immediately.

4. Discussion

4.1. Result Considerations

The integrated demonstration balancing tool shows promising results. The demonstrator provides the possibility to manually optimize musculoskeletal risks and timeline balancing in parallel, a relatively unique functionality according to Neumann and Dul [3]. The demonstrator offers a good potential to move away from the traditional approach, where assembly time and production capacity are the dominant parameters, to an approach where human well-being is considered equally important and taken into account before production equipment and other investments restrict possible later needed modifications of the solutions. The suggested approach shows that it has a positive impact on human well-being and does not necessarily reduce productivity. The time to perform the two more complex parameter optimizations in parallel (Test 2:1) was shorter compared to the serial process, where time was first balanced and afterwards an adjustment was made according to advice from an ergonomic expert (Tests 1:1 and 1:2). Most of the test subjects had a positive attitude toward the software tool used in Experiment 2, which indicates that the usability of the demonstrator seems acceptable.

The presented balancing demonstrator includes manual multi-objective optimization, which could be automatized in future. However, modifying the manual balancing process currently used in industry by adding functionality for manual multi-objective optimization of productivity and well-being is a small modification that users in industry will probably accept which is also indicated in the follow-up questionnaire. More advanced balancing approaches, based on fully automatic optimization processes, according to Falkenauer [19], have not yet been implemented in industry settings. The industry trend is going in that direction, using big data and automatic processes for supporting decision making, but industry does not yet seem to be ready. It may take some years before industry is ready, and the presented demonstrator is a constructive mid-step in the direction of ongoing industry transformation. It is also one step toward the consideration of ergonomics in industry in accordance with the definition of ergonomics, i.e., optimizing well-being and overall system performance [2].

4.2. Method Considerations

There are many different process planning simulation tools used for balancing, such as Siemens^TM, Dassault^TM and AVIX^TM. Furthermore, there are also several different methodologies for estimating ergonomic load, e.g., academic methods, such as OWAS [28], OCRA [17], RULA [29], and EAWS [30], as well as company-specific assessment methods. In addition, there are several different time-setting methods, MTM, MTM-UAS, MTM-SAM, and clocked time [31]. The demonstrator uses a selection of existing software and human well-being methods; the concept was demonstrated using clocked time, AVIX, and REBA. However, the concept is believed to work with any software and method combination. A key for the presented concept to be implemented in industry is most likely to base the solution on the balancing tool and the human well-being assessment method already used at the specific company, i.e., to improve and merge the existing tools and methods instead of introducing new tools and methods.

The process of comparing the current balancing process (Tests 1:1 and 1:2) with the proposed concept (Test 2:1) was structured, and all test subjects made the test in the same order. Firstly, the cycle time was optimized with musculoskeletal risks indirectly considered based on the subjects’ knowledge. Secondly, the subjects modified the balance solution from Test 1:1 based on feedback from an ergonomist on the worst stations from a musculoskeletal risks point of view. Lastly, in Test 2:1 the subjects optimized cycle time and musculoskeletal risks in parallel. Such a set-up may have had a learning effect. An alternative method could have been that subjects randomly started with either Experiment 1 (Tests 1:1 and 1:2) or Experiment 2 (Test 2:1). However, knowledge obtained about ergonomic assessments of case-specific tasks in Experiment 2 might influence results in Experiment 1. Furthermore, Experiment 2 was carried out several weeks after Experiment 1, which may have largely reduced the possible learning effect.

4.3. Future Research

Given the promising results, the user-centered development of the concept should continue. A logical next step is to replace REBA with the company-specific method, followed by a broader introduction and evaluation of the tools among the automotive industry partners involved in the project. Another promising direction would be to automate the process and conduct a multi-objective optimization that accounts for both musculoskeletal risk and cycle time.

5. Conclusions

The findings indicate that the proposed manual parallel optimization approach, designed to address both musculoskeletal risk and cycle time, is not only well-received by the test subjects, but contributes positively to their overall production preparation work. Through the application of this method, several solutions were identified that effectively lowered musculoskeletal strain without compromising productivity, as measured by cycle time. Moreover, the process of conducting the more complex two-parameter optimization in parallel demonstrated a substantial reduction in time compared to the traditional sequential method. These results suggest that integrating a user-centered, parallel optimization strategy can enhance both ergonomic conditions and efficiency in industrial settings.