The work balance chart is a critical component of a line balancing effort. It is both the graphical representation of the allocation of task time among operators, equipment, and transfers in a manufacturing or service process and a tool used to achieve an equal distribution. Like other tools discussed in “The Third Degree,” a work balance chart may be referenced by other names in the myriad resources available. It is often called an operator balance chart, a valid moniker if only manual tasks are considered. It is also known as a Yamazumi Board. “Yamazumi” is Japanese for “stack up;” this term immediately makes sense when an example chart is seen, but requires an explanation to every non-Japanese speaker one encounters. Throughout the following presentation, “work balance chart,” or “WBC,” is used to refer to this tool and visual aid. This term is the most intuitive and characterizes the tool’s versatility in analyzing various forms of “work.” A work balance chart can be used to streamline manufacturing or service operations. Any repetitive process that involves manual operations, automation, and transfers of work products between operators or equipment can benefit from work balance analysis. Applications in manufacturing are most common; service providers should take heed of the potential competitive advantage such an analysis could bring to light. Assumptions To simplify the initial presentation, preparation for and construction of a work balance chart is subject to several assumptions, including:
Preparation Before an analysis can begin, its boundaries must be established. Define the process completely, including its start point, end point, and each step between. Creating a flow chart is a convenient way to document the process definition for reference during analysis. Though known, constant activity times are assumed, this information must be collected and organized. If not already available, a precedence diagram should be created; referencing preexisting diagrams is much more efficient than creating them “on the fly” or, worse, foregoing them (“running blind”). Activity times can be added to reference diagrams to reduce the number of documents needed during analysis. Summing the activity times for every step in the process yields the total cycle time or TCT. Determine the takt time – the maximum cycle time at which the process is capable of meeting customer demand. Takt time is calculated by dividing the available work time in a period by the customer demand in that period: For manufacturing operations, the shipping frequency provides a convenient period for takt time calculation. In this framing, The period used for service operations is often a shift, a day, or a week, depending on the duration of the service analyzed. It is possible, however, that another, “nonstandard” time period is appropriate; its selection is left to the judgment of the analysis team. To determine the optimum number of operators, or stations, for the idealized process, divide the total cycle time by the takt time and round up to the nearest integer: Constructing the WBC A work balance chart, like several other diagrams, can be constructed manually or digitally. Many diagramming efforts benefit significantly from manual construction; speed of development, the cross-section of inputs, and size of the workspace (“canvas”) are typical advantages. The nature of a WBC, however, often shifts the advantage to digital construction for experienced practitioners, as a description of both methods illuminates. Manual construction of a WBC is most valuable for training purposes; advantages in this context include:
On a piece of paper, draw a graph with an appropriate time scale, chosen based on the takt time and task times, on the vertical axis. Along the horizontal axis, place labels for each station or process in sequence from left to right. For each task, cut a strip of paper to length in proportion to its duration; use the scale established on the graph. To improve legibility and maintain the proper scale, the graph and task strips can be printed on a computer; these will look similar to the example in Exhibit 1. Doing so reduces the time required, maintaining focus on the important aspects of the WBC; the ability to draw time scales accurately is the least relevant to a work balancing effort or training. On the blank graph, draw a horizontal line at the takt time or target cycle time (they may be different – more on this later). Use a bright or contrasting color to ensure that the line is easily visible. Place the task strips on the graph, “stacking” them above the corresponding station or process label. An example of what this might look like is shown in Exhibit 2. The example presents the current state of production that clearly exhibits a huge imbalance in the workload. To balance the workload among the available stations, rearrange the task strips, targeting equal total task times in each station. Each task movement is subject to restrictions established by the precedence diagram for the process being analyzed. The future state of production, with a balanced workload, may look like that presented in Exhibit 3. A task eligibility chart can also be created; in it, information contained in the precedence diagram or table is reorganized to be more easily applied directly to the work balance effort. To see how an eligibility chart is created and utilized, consider the example precedence diagram and table in Exhibit 4 and the derived eligibility chart in Exhibit 5. In this example, there are 12 tasks to be completed to meet a 57.6 s takt time (the presumed target cycle time). The number of stations needed is determined by dividing the total task time by the takt time: 252 s/57.6 s = 4.375. Rounding up, the work is to be balanced among five stations. For a task to be eligible for assignment to a station, it must meet all precedence requirements. A commonly used “rule of thumb” is to assign the eligible task with the longest duration that will not cause the total station time to exceed the target cycle time. In the example presented, the tasks are assigned in alphabetical order, but this need not be the case; a process with more parallel tasks will have more “mixing” of the task sequences. The work balance chart in Exhibit 6 provides the graphical representation of the eligibility chart assignment information. The WBC in Exhibit 6 was generated in a spreadsheet program; compare it to Exhibit 3. The output of the manual (hybrid, really) process depicted in Exhibit 3 is functional, but imprecise and aesthetically unsatisfying for presentation. This is true despite the use of a computer to generate the graph and task strips. With comparable effort, a spreadsheet template can be created to generate an aesthetically pleasing WBC that automatically adapts to task rearrangement. In subsequent balancing efforts, use of the template is much more efficient than the manual process. For this reason, experienced practitioners are encouraged to construct WBCs digitally. Once the concepts of proper execution are well-understood, physical manipulation no longer adds sufficient value to justify an inefficient process. A spreadsheet template can also be created to generate a horizontally “stacked” WBC, such as that shown in Exhibit 7 for the eligibility example. Exhibit 6 and Exhibit 7 present the same information in slightly different formats and with different connotations. The vertical “stacks” of Exhibit 6 may evoke the concept of “piling on” or putting an operator under increasing load as additional tasks are assigned to a station. The horizontal bars of Exhibit 7 tend to be less evocative, portraying the inevitable, and mostly unobjectionable, passage of time. The notions of “workload” and “timeline,” though equivalent in this context, can elicit very different reactions from varying audiences. Both formats provide accurate, acceptable presentations of the work balance; the choice between them is made for aesthetic reasons. Adjustments for Reality The examples in the previous section allude to use of work balance charts to improve an existing process (see Exhibits 2 and 3) and for process planning (see Exhibits 4, 5, and 6). The information that can be known and that which must be estimated differs between these two applications. For example, data from a time study can be used to balance an existing process, but task times must be estimated in a preliminary process plan (i.e. prior to building equipment). Once performance data is available for a new process, the workload may require rebalancing. This is often done assuming a constant activity (task) time; the average of recorded cycles is typically used. However, this may not be the most effective choice; an “accordion effect” of varying task durations may induce erratic fluctuations in the workflow. These fluctuations can be accommodated in the target cycle time. Activity times are typically normally distributed; this can be verified in the performance data prior to implementing the following strategy. Consider a hypothetical task for which time study data reveal an average duration of 30 s and standard deviation of 4 s [μ = 30, σ = 4]. To smooth the workflow, the task time “standard” is set to encompass 90% of cycles. As the normal distribution for this example, Exhibit 8, shows, this task duration is set at 35.1 s [P(x < 0.9); z = 1.282]. Variation in activity time is unavoidable for manual tasks; thus wait time (waste) will exist in a closely-coupled process. Monitoring and evaluation of productivity and operators’ frustration with the system is required to choose an appropriate target cycle time. Designing a system to operate at the average task time ensures that only 50% of cycles will meet the design criterion! The system output will meet design intent only when the average task time is achieved at every step of the process. This tension is relieved if physical buffers are built into the system or output from stations is batched; these practices have an equivalent effect. Variation in task time does not influence downstream operations; the average task time is the most relevant metric for most systems of this type. Takt time may also vary; this is the nature of seasonal demand, for example. Periods of reduced demand can be accommodated in a few ways:
Several of the assumptions upon which the WBC development presentation was based are interconnected; it is difficult to discuss one without invoking others. The assumption that “target cycle time equals takt time” may be removed for many reasons, some of which were presented in other assumptions. If resources – personnel, equipment, etc. – are shared with another process, the target cycle time may be reduced so that sufficient time is available to meet the demand for both processes. This creates a situation similar to introducing product mix into a process, precluded from this discussion by another assumption. This topic is best left to a future installment; adequate exploration is beyond this scope of this presentation. The assumptions of “100% process availability” and “100% acceptable quality,” like that of constant activity times, were made to avoid confusing those new to line balancing. They must now be adjusted, however, to develop realistic expectations of process performance. Quality and availability are two legs of the OEE (Overall Equipment Effectiveness) “stool.” Achieving 100% performance in both measures for an extended period of time is unlikely for a system of any sophistication. Therefore, the target cycle time must be adjusted to accommodate the actual or anticipated performance of the system. The reliability of a system effects its availability and directly influences the numerator of the takt time calculation. Output of unacceptable quality is accounted for, indirectly, in the denominator by effectively increasing the demand (an additional unit must be produced to replace each faulty unit). The modified calculation can be presented as: The third leg of the OEE stool is productivity, to which the task time variation discussion and Exhibit 8 allude. An example occurrence of these three losses is depicted in Exhibit 9. Adjusting target cycle time based on OEE is an alternative method (“shortcut”) of compensating for these losses when they are known or can be estimated with reasonable accuracy. For this method, the target cycle time is calculated as follows: These calculations are, once again, based on the assumption that resources are dedicated to a single process. For process planning, the target cycle time can be calculated using a “world-class” OEE of 85%, an industry average, or other reasonable expectation. The final assumption stated involves continuous improvement (CI) efforts. Processes evolve to accommodate changing customer requirements, material availability, and other influences on production. Many times, process changes are implemented with incomplete analysis, whether due to urgency or oversight, resulting in a system that is inefficient and unbalanced. Learning curves may also effect tasks differently; experience may improve performance in some tasks more than others. Including this assumption in the discussion serves as a reminder that line balancing is a CI effort involving both efficiency and arrangement.
Additional Notes on Balancing If no satisfactory balance can be found without exceeding the target cycle time, there are several approaches available, including:
The digitally-generated WBC examples (Exhibits 6 and 7) were created with “traditional” spreadsheet data presentation and charting tools. Pivot tables can also be used to organize data and generate charts; however, their use requires additional skills and manual updates of charts. If one is sufficiently skilled and comfortable in the use of pivot tables, it is a viable option, though the advanced users that benefit from their use is probably a small fraction of practitioners. Notes on Simulation Simulation software can be used to facilitate line balancing and other operational assessments. Work balancing projects like those described here will usually not benefit greatly from the additional effort that simulation requires; highly sophisticated systems may warrant it, however. A complex product mix, highly variable task durations, complex maintenance schedules, and unpredictable demand may complicate the analysis to a sufficient degree to justify the use of simulation software. A spreadsheet program can also be useful for “what if” type experimentation and is sufficient for most line balancing projects. Monte Carlo simulation, distribution analysis, and other simple functions can also be performed in a spreadsheet. Approaching the limits of production capability requires the most complete and accurate information possible. It is imperative to account for variability in human task performance, equipment reliability, quality attainment, predictability of demand, and other factors in process planning and development. Increasing efficiency and improved work balance are circular – each supports the other – and should be pursued in conjunction whenever feasible. For additional guidance or assistance with line balancing, or other Operations challenges, feel free to leave a comment, contact JayWink Solutions, or schedule an appointment. For a directory of “Commercial Cartography” volumes on “The Third Degree,” see Vol. I: An Introduction to Business Mapping (25Sep2019). References [Link] Line Balancing Series. Christoph Roser. All About Lean, January 26, 2016. [Link] “The Balancing Act: An Example of Line Balancing.” Brian Harrington. Simul8. [Link] “Operator Balance Chart.” Lean Enterprise Institute. [Link] “Understanding the Yamazumi Chart.” OpEx Learning Team; July 19, 2018. [Link] “What Is Line Balancing & How To Achieve It.” Tulip. [Link] “Lean Line Balancing in the IT Sector.” Rupesh Lochan. March 9, 2011; iSixSigma. [Link] Normal Distribution Generator. Matt Bognar. University of Iowa, 2021. [Link] Normal Distributions. [Link] The New Lean pocket Guide XL. Don Tapping; MCS Media, Inc., 2006. [Link] The Lean 3P Advantage. Allan R. Coletta; CRC Press, 2012. Jody W. Phelps, MSc, PMP®, MBA Principal Consultant JayWink Solutions, LLC jody@jaywink.com
0 Comments
Your comment will be posted after it is approved.
Leave a Reply. |
AuthorIf you'd like to contribute to this blog, please email jay@jaywink.com with your suggestions. Archives
November 2023
Categories
All
![]() © JayWink Solutions, LLC
|