Simulation is a valuable tool for lean Six Sigma
By Paul D. Babin and Allen G. Greenwood
Discrete-event simulation is a powerful computer modeling tool that helps engineers understand the impact that variability can have on a production system. Modeling a proposed change in a system during the planning phase of an improvement project greatly increases the likelihood and confidence that the improvement will generate its intended benefits. Six Sigma black belts and lean project team leaders should understand when to use discrete-event simulation on their projects and where to go for help in developing this skill set for their teams.
Simulation works well with other techniques and is suited for understanding variability in complex production systems. Those systems can exist in a wide variety of domains besides manufacturing, including logistics and service operations. Simulation can be applied at various levels, from the micro (say, the study of operator-machine interactions in a manufacturing cell) to the macro (such as an entire supply chain).
Six Sigma projects allow for drawing a critical to quality (CTQ) tree to help understand how process steps translate to important customer features. Other cause-and-effect analysis tools are used to understand which factors might have the biggest impact on output variability. These help form a hypothesized model of a system, y = f(x). The quest is to determine which inputs (the x’s) to the system have the largest impact on the behavior of the system. This is reflected in outputs (the y’s).
But what if the system is complex, and there is uncertainty as to the effect of input variability on the output of the system? Also, what if the relationships between the inputs and outputs of the system are highly interrelated and nonlinear? In these cases, simulation can really help with Six Sigma projects.
Lean is focused on improving productivity by eliminating waste in all of its forms. This starts with defining value from the customer’s perspective and defining waste as anything that does not add value. A flow of value through the system is created, and work continues to perfect the process in order to achieve long-term benefits.
In lean production projects, improving the flow of value could include eliminating excessive transportation steps, reducing work-in-process inventory and moving from a push to a pull system. For simple systems, it is not difficult to visualize the improvement and to use estimates of steady-state averages. But when the system is complex and details beyond just long-run averages are needed, simulation can help.
THE RIGHT STUFF
Simulation builds collective understanding of the process being improved. It has eight primary advantages when used during a process improvement project:
- Brings structure
- Requires quantification
- Uses the best measures
- Opens the decision space
- Promotes experimentation
- Enables the understanding of impact of variability
- Promotes taking a systems view
- Encourages collaboration
What is discrete-event simulation?
There are various types of simulation. Many engineers might be familiar with Monte Carlo simulation. If a system is represented in a spreadsheet, and if there is uncertainty about some of the parameters, then Monte Carlo simulation is used to examine the sensitivity of the system’s performance to that uncertainty. Such simulation can generate distributions of outcomes or performance measures based on the uncertain inputs. For example, in a financial model, Monte Carlo simulation can be used to see how likely it is for the return on investment to exceed a threshold based on uncertain cash flows and interest rates.
Discrete-event simulation goes beyond Monte Carlo simulation when the model considers interrelated events that occur over time. These events affect various states in the system that give rise to behaviors and performance. Discrete-event simulation is useful in queuing and inventory systems where dynamics make the system hard to analyze with a spreadsheet. Discrete-event simulation is necessary when there are complex routings, including a sequence of process steps that might be conditional based on results of a previous operation. For example, suppose that 10 percent of the time a product has to go to a rework station to wait for service. How will that affect performance?
Both of these simulation modeling approaches enable predictions about how a system will perform in the future, facilitating complex “what-if” analyses. Both approaches extensively use random variables. In discrete-event simulation, for example, random variables represent time between arrivals, process times, whether an item is of acceptable quality, and various item characteristics.
Basic elements of discrete-event simulation include constructs that represent arrival processes (random or scheduled), items flowing between activities based on various routing logics, and activities that process the items by using fixed and mobile resources. The resources may incur planned and random downtime. Discrete-event simulation processes a large number and wide variety of related events over time and samples from a range of probability distributions to capture the variability in a system’s behavior. All major commercial simulation modeling packages include these aspects and have advanced or specialized elements to make the model more realistic. The packages often allow users to customize elements and logic through a built-in programming language.
Commercial simulation software also typically includes a means to carry out experiments with simulation models, including managing replications and scenarios, performing statistical analyses and summarizing results. Most simulation software has some level of animation capability so stakeholders can “see” what happens as the model runs. This is useful when presenting results to a nontechnical audience. Animation also helps validate, verify and debug the model to make sure it behaves the way it is intended.
Commercial software packages have made simulation models easier to build and use. No matter which package is chosen, simulation can enhance the process improvement process.
The simulation project approach
A typical approach for using simulation in a project is illustrated in Figure 1. First, define the objective, scope and requirements of the improvement opportunity. Know how the simulation model will be used before it is designed and built. Involve all of the stakeholders in the defining and planning process. This definition step leads to three parallel activities – formulating the model, preparing the data and designing the experiments that will be conducted using the model.
Simulation models are data driven; therefore, it is important to identify the data needs early in the project since data collection may be extensive. The data must be reviewed and analyzed, such as fitting varying process times to a representative probability distribution before it can be used in the simulation model. Often the data for the model parameters may not exist and will need to be estimated. Many techniques used in the “measure” step of the Six Sigma DMAIC project approach apply here.
To compare improvement alternatives, you must identify the input or decision variables that will be changed. Similarly, you must define the basis, or performance measures, that will be used to compare the options. It is important to address experimentation early in the process so the model includes the decision variables and can estimate the performance measures.
The simulation model of a system is best built in two major steps – developing a conceptual model and translating that representation into a computer model. The conceptual model is a nontechnical diagram of the logic and flow, similar to a value stream map, and ensures that key elements and their relationships provide a valid representation of the system. The conceptual model then is translated into a computer model using special-purpose commercial simulation software. Significant skill is required for model building, but if the requirements and system are well-defined, the model-building step becomes more straightforward.
Once the model is built and validated, numerous experiments can be run to predict how the system will perform under a variety of conditions.
Finally, simulation results are presented to the stakeholders, including managers, process owners and users.
Lean and Six Sigma are all about trying to get more improvements from production systems. If we consider lean to focus on reducing waste and Six Sigma to focus on reducing variability, then we consider simulation to focus on assessing the impacts of proposed changes in those improvement approaches.
REAL BENEFITS TO SIMULATION
- Simulation can enhance lean Six Sigma efforts by providing a means to test the impacts of possible changes before they are selected and implemented. Animation built into simulation allows everyone to “visualize” the dynamic behavior of possible lean systems.
- For complex systems, simulation may be the best way to quantify the improvement opportunity.
- Simulation is a great tool for understanding and experimenting with ways to reduce variability in complex production systems.
- Simulation helps manage the inherent uncertainty in production systems by seeing how the system responds to a full range of possible system inputs.
- Simulation fosters taking a systems view; this is important since it can help one understand the effect of a proposal on upstream and downstream operations and its effect when combined with other improvements.
Advantages of using a simulation approach
There are eight primary advantages for using a simulation as part of a process improvement project.
1. It brings structure to the process. To build a simulation model, the step-by-step process elements must be defined. How items flow must be understood. It also must be clear what information is to be obtained from the model. The basis for comparing alternatives, or the performance measures, must be specified.
2. Simulation requires quantifying parameters through data analysis or estimation and specifying probability distributions. For example, processing times need to be specified as distributions, either by collecting data – doing time studies or going to the database – or through engineering estimates.
3. Simulation can use the “best,” most appropriate, measures of system performance. In many process improvement projects, the desired performance measures are not available. Thus, we settle for other measures because they are available, for example, in the accounting system. While it may not be the “right” measure, it is the only one we can get. In simulation, any performance measure can be defined. To do so, a variable is defined and added to the model to track the measure. Therefore, simulation encourages thought as to what should be measured.
4. Simulation opens up the decision space. It provides a quick and inexpensive way to consider a range of options and identify the effects of proposed changes on system performance. The alternatives become changes in the model and not disruptions to the real system. For example, in order to remove a bottleneck, what is the impact of adding another machine, working an extra shift, improving maintenance? Simulation also provides a valuable means to identify where a bottleneck moves to once the current constraint has been relaxed.
5. Simulation promotes experimentation. The simulation model can be used to conduct sophisticated “what-if” analyses. Alternatives can be tested under a variety of conditions in order to assess their robustness. For example, what is the effect on lead-times if sales increase 10 percent, 20 percent or 50 percent? And since simulation models run much faster than real time, long-term effects of changes can be analyzed easily.
6. Simulation is often the “best” tool to understand and quantify the impact of variability on a system. Since variability corrupts, it detracts from productive capacity. Simulation, unlike any other tool, facilitates the examination and analysis of the interaction of multiple random processes. It also provides details beyond just the averages, such as measuring the percentage of customers who have to wait for a longer time than an important threshold.
7. Simulation promotes a systems view because it can be used to understand a proposed change’s upstream and downstream effects. It is not uncommon for an improvement in one area to have a negative effect elsewhere in the enterprise. Simulation can identify those conflicts and select what is best overall, not just what is best for the area being studied. This avoids sub-optimization. Simulation can be used to examine the interaction of multiple concurrent improvements that are being considered.
8. Simulation encourages collaboration. At the beginning of a project, all the stakeholders should agree on the objective, decision variables and performance measures. Throughout the project, and especially during the simulation model-building phase, stakeholders should contribute to defining the system, providing data and other inputs, validating the model and interpreting the output.
Therefore, the use of simulation leads to a collective understanding of the process that is to be improved and which input variables have the largest impact on performance.
When to use simulation
Ideally, simulation is employed during the project planning stage. This increases the confidence in the degree to which the change will deliver the anticipated results. Since simulation requires a solid understanding of the system and the planned changes, and since it can quickly evaluate the change under a variety of conditions, simulation reduces project risk.
Simulation is a means to model complexity. If a process being studied is quite simple, then a simulation model may not be needed. Simulation becomes an important tool for a more complex system with many elements, a large amount of interdependencies among the elements, and/or considerable variability.
Simulation is descriptive rather than prescriptive; it defines how a system works, not what’s best. Thus, simulation describes how a system likely will react to proposed changes, not how to configure the system for optimal results – at least not by itself. For example, simulation can estimate system cycle time if another workstation is added, but not the optimal number of workstations. In that case, several scenarios would need to be run and the best performer selected. Many simulation packages work with optimization packages so that together they can find the optimum solution.
Simulation models provide a great means to communicate with stakeholders. The groups that the project will affect can see what it will look like and how it will operate and behave when implemented. Simulation with animation is much more effective than two-dimensional static drawings to build understanding and confidence.
The Six Sigma world has many project opportunities and many people trained in Six Sigma. The key problem, though, is what projects should be worked on so as to use resources most effectively. Simulation can help identify which projects will have the largest overall impact.
Simulation need not be used if the answer to a problem is obvious. Simulation also isn’t needed for small projects with small risk and short time periods for completion. Sometimes, the effect of a change, even a simple change, is uncertain. Simulation can help improve our confidence. It may be worth it not to jeopardize the reputation of the improvement approach because of an unintended system consequence that could be discovered through simulation.
Don’t skip simulation just because you are not skilled with the tool set. Invest in obtaining the capabilities needed to perform simulations when needed. Have someone in your group who has this skill and experience to call on when needed.
There are many ways to get started with simulation. One way is to take a class. Courses are offered by simulation software companies and simulation consultants, as well as by industrial engineering programs at many universities either through academic departments or outreach services. Most recent industrial engineering undergraduates have been required to take at least one course in simulation. Similarly, there are a number of places to look for assistance once you get started. Many simulation software companies offer assistance with on-site training or project consulting to get started. Consultants can help with larger projects.
Your first simulation projects should be small and relatively easy. While simulation helps large, complex projects in particular, it is wise to develop the skill set on a small starter project. That way, the organization gains confidence and the modelers gain credibility.
Finally, like any tool, simulation needs to be kept sharp. There are a number of simulation sessions at the IIE Annual Conference & Expo in May. IIE also co-sponsors the annual Winter Simulation Conference, which provides a wide range of tutorials at varying levels, research papers, case studies and vendor exhibits. The WSC proceedings are available online at www.wintersim.org.
Simulation modeling can be a powerful tool in your continuous improvement toolbox.
Paul Babin recently joined ThyssenKrupp Elevator as senior director of quality and operations excellence. He holds a B.S. in electrical engineering and an M.B.A. from Christian Brothers University, an M.S. in electrical engineering from the University of Tennessee, and an M.S. in industrial and systems engineering from the University of Memphis. He is working on his Ph.D. in industrial engineering at Mississippi State University. Babin has worked in industry for more than 25 years, including for Texas Instruments, Dover Elevator and Mueller Industries, and he ran his own engineering consulting company. He is a senior member of IIE.
Allen Greenwood is a professor of industrial and systems engineering at Mississippi State University. He received his B.S. in industrial engineering from North Carolina State University, his M.S. in industrial engineering from the University of Tennessee, and his Ph.D. in management science from Virginia Tech. Greenwood spent nine years in industry, including work with American Enka and General Dynamics Corp. Prior to joining the Mississippi State University faculty in 1994, he was on the faculty at Virginia Tech, Northeastern University and the American University of Armenia. Greenwood is a senior member of IIE.