What should a process workflow look like? It should reflect the product or ser- vice customers need and value. In manufacturing this concept is embedded within the bill of material (BOM) hierarchy as well as design documentation such as the DFMEA. A BOM is used as the basis on which to design a production system because it shows how a product is constructed at every level of its design. The BOM also shows hierarchal relationships between each level of the design. As a simple example, an automobile has four wheels and each tire has five lug nuts. The BOM associated with this product would include four wheels per automobile and five lug nuts per wheel, each related in a hierarchal manner. This concept also applies to service industries. As an example, McDonald’s builds hamburgers. The BOM of a hamburger would specify one roll split in half, one hamburger patty and pickles, as well as other materials to be placed on the hamburger. The engineers designing McDonald’s process would build its workflows to include this BOM information. Finally, in the case of financial service organizations, customers who purchase vari- ous mutual funds and other financial services would have a product and service portfolio (BOM) created for them to provide their financial advisor with informa- tion useful in managing their portfolios.
Table 5.3 lists 20 basic steps necessary to ensure a new product or service can be successfully produced to meet organizational goals and objectives. In addition to the information listed in Table 5.3, process engineers should obtain additional information from the CE team. Several of the 20 steps require documentation from the design phase of the project. This documentation includes a preliminary process flowchart based on the product’s BOM, or the service package in the case of a new service. The preliminary process flow diagram is developed by process engineers
Operational Flexibility
1. VOC Driven Product and Process Designs 2. Application of Design-for-Manufacturing Methods 3. Value Stream Map (Eliminate Steps and Operations) 4. Manage Bottlenecks (Maximize Flow)
5. Mixed Model Operations (Minimize Set-Ups) 6. Transfer Batches (Direct Lead Time Reduction)
7. Apply Lean, Just-in-Time (JIT) and Quick Response Methods (QRM).
Variety
Low High
Low High
Product Volume
Using Lean Methods to Design for Process Excellence n 137
who were CE team members. It should include spatial relationships as well as the key inputs and outputs of every operation. To ensure the new design has been fully evaluated relative to its failure points, a DFMEA and preliminary quality control plan should also be available to the process engineers. Process engineering also works with suppliers to design new equipment, tooling, and facilities (if necessary), as well as measurement and testing equipment. A PFMEA should also be created by process engineering as part of its prototyping. The PFMEA is comparable to the DFMEA except that process workflows are aligned with the critical design charac- teristics at each process step. The prelaunch control plan is built using information
table 5.3 20 key Steps to Create new Product or Service Workflows
1. Preliminary process flowchart 2. Product assurance plan
3. Design failure-mode-and-effects-analysis (DFMEA) 4. Preliminary quality control plan
5. New equipment, tooling, and facilities requirements 6. Gauges/testing equipment requirements
7. Product/process quality systems review 8. Process flowchart
9. Floor plan layout
10. Process failure-mode-and-effects-analysis (PFMEA) 11. Prelaunch control plan
12. Process instructions
13. Measurements systems analysis 14. Production trial run
15. Preliminary process capability study 16. Production part approval
17. Production validation testing 18. Packaging evaluation 19. Production control plan
from the DFMEA and PFMEA to communicate to the organization important information that is necessary to produce the product or service. Integral to the final process workflow design is creation of work instructions, measurement systems, preliminary process capability studies, and related documentation. This informa- tion is used to plan the production trials, whose purpose is to collect samples or other information related to the performance of the new product or service for anal- ysis by design and process engineering. These production validation activities are used to ensure that customer requirements are met by the new product or service under actual production conditions within the new process workflow. If produc- tion validation testing is successful, the new production system will be scaled up for full commercialization. In parallel, various components of the packaging design are evaluated to ensure they also meet customer requirements. Finally, the production control plan and all related documentation are updated across all organizational functions. The outputs from these integrated activities are a finalized process flow- chart, work and inspection procedures, floor plans, and a project schedule to scale up and commercialize a new product or service. The project schedule includes the balance of the deliverables that are necessary to support the new process workflow.
Workflow modeling
Developing a capability to model your process workflows allows experimentation of the workflow model offline as well as evaluation of numerous alternative workflow design concepts under varying modeling conditions. But, too often, organizations do not use workflow modeling methodology due to resource limitations, organi- zational barriers, or a lack of training. The ten key steps as shown in Table 5.4 will allow an organization to successfully implement workflow modeling. The first step is to bring together a group of people who have been trained to build and ana- lyze workflow simulation models. This group should have a strong background in engineering or statistics and preferably at a graduate level. Training provided by workflow software vendors may also be very useful to the team. The second step is to develop a list of applications where the model’s methodology can be realistically applied within your organization to create benefits such as reductions in time or cost, or simplification of the workflow design. To facilitate this work, several common modeling methodologies can be used by a team, depending on the type of workflow and questions that need to be answered by the simulation analysis. These methods include simulation, queuing analysis, linear programming, and customized models and algorithms that have been developed for very specific applications.
Once the team has determined the type of required modeling methodology, it can begin the process of researching off-the-shelf software and hardware. Modeling software has been developed for a wide range of applications based on differing workflow assumptions. These assumptions may range from simple to complex. The software that models simpler workflows can be generic and applied to diverse process
Using Lean Methods to Design for Process Excellence n 139
applications, but to the extent generic software models must be customized, they may require much effort to set up and analyze. It is always more efficient to purchase software that has been designed specifically for your workflow application because it will be easier to develop and interpret the model. Also, training will be easier using specialized software. As an example, there are many different types of off-the- shelf modeling software for workflows such as manufacturing processes, financial services, call centers, warehousing, inventory management, distribution networks, and many others. It makes sense to purchase the type of workflow software that is easily configurable to system elements that reflect your workflow rather than create workflow models from the ground up. As part of the process of researching software systems, a modeling team should develop a library of examples that show how their organization’s workflows will be modeled. This will help the team communicate the advantages of workflow modeling to its organization.
After the modeling team has selected the necessary software and hardware, it begins to build workflow models. These activities include developing the under- lying structure of the workflow model, including its goals and objectives, system constraints, and parameter settings. In addition, probability distributions are
table 5.4 10 key Steps to model Workflows
1. Organize a group of people who have been trained to build simulation models.
2. Develop a list of areas in which the model’s methodology can be realistically focused.
3. Research and select off-the-shelf modeling software and associated hardware to match expected process applications, i.e., manufacturing, service systems, warehousing, logistics, etc.
4. Develop a library of applications that can be used as examples of applying the model within your process.
5. Develop the underlying model structure, including its goals and objectives, system constraints, and parameter settings.
6. Determine probability distributions and time span of the model. 7. Develop decision rules, including initial and final states of the model. 8. Develop plans to obtain the necessary process data to test the model’s
accuracy.
9. Analyze the output of the model using statistical tests to determine significance of the model’s output.
10. Document and communicate the model’s results, and develop plans to implement solutions as practical.
determined and integrated relative to the model’s inputs and outputs for the vari- ous steps or operations within the workflow. The model’s underlying structure and form, the time span of the analysis, and its decision rules are also developed to mirror the key characteristics of the real workflow, including its initial and final states. After the workflow model has been created, the team obtains the neces- sary process data to test the model’s accuracy. In these evaluation activities, the performance characteristics of various workflows are evaluated under simulated conditions. Finally, the various analyses are documented in an appropriate format and communicated to the larger organization in the form of recommendations and practical solutions to improve process workflow efficiencies.
Figure 5.5 shows that workflow models are virtual representations and should correspond to actual characteristics of the process workflow. In real-world situa- tions, the underlying interrelationships of the numerous workflows within a process may initially be unknown or poorly understood by an organization. In particular, the dynamic and complex performance characteristics of workflows cannot usually be fully understood in terms of their theoretical or actual performance without
Forecast Operation Schedule Variation Due to Process Breakdowns Demand Variation Variation in Labor, Materials and Capacity Schedule Met? Real World Process Characteristics
•Unknown Process Performance and Inter-Relationships •Dynamic and Complex Performance of System Components •Ambiguity and Poor Resolution of Performance
•Time Delays Between Events and Their Measurement
•Mapping of Real World Characteristics Using Basic Process Characteristics, Parameters and Decision Rules.
Virtual World Process Characteristics
•Known Model Performance
•Specified inter-Relationships Between System Components •Ability to Experiment and Resolve Inter-Relationships •Ability to Compress Time Between Events
Using Lean Methods to Design for Process Excellence n 141
creating models. This is because operational relationships are complicated and not obvious to a casual observer of the process workflow. Also, actual systems are char- acterized by ambiguity and poor resolution of performance, as well as time delays between the occurrences of events and when events are actually measured by an observer. This makes it difficult to understand the relationships between causes and their effects in actual workflows. Thus, mapping of actual workflow charac- teristics, including their parameters and decision rules, into a virtual model of the workflow is very useful in improving operational performance. The advantages of using a workflow model are that its structure and the dynamic performance of its operational components are known by an analyst. Also, an ability to experiment on a system and compress the time between workflow events enables an analyst to evaluate numerous alternative workflow designs and identify an optimized version of the process workflow.
Simulation
Simulation methods are useful in modeling a diverse set of process applications. These include analyzing system capacity under various constraints, comparing system performance among several alternative workflow designs, and conducting sensitivity analyses to determine the impact on a system’s outputs relative to varia- tions of one or more key process input variables. The major advantage of using a simulation model is that it can be designed in a flexible manner to model a work- flow, and its event probabilities can be easily defined by their underlying probability distributions. On the other hand, if your model is goal oriented, such as maximiza- tion or minimization of an objective function, and subject to clearly defined and deterministic constraints, then linear programming models might be more useful in modeling a workflow. Queuing models provide additional tools and methods to model process workflows, but the process workflow must fit predefined criteria. We will discuss the application of linear programming and queuing models to work- flow analysis later in this chapter.
The first step in creating a simulation model requires identifying the model’s goals and objectives as well as its structure and constraints. Specifically, the team should ask: Why is the simulation model being created, and what are the expecta- tions of the analysis by the process owner of the workflow? Other relevant consider- ations may include the simulation project’s budget and schedule. It should be noted that developing simulation models usually takes longer than expected because the model must be modified in an iterative manner to ensure its performance corre- sponds with that of the actual workflow. The second step in developing a simulation model requires defining the scope of the project relative to the workflows modeled by the team. In other words, where does the model begin and end relative to the workflows under evaluation? The third important step is identifying the underlying functional relationships, i.e., Y = f(X). These functional relationships are the basis
of the simulation model. In particular, the relationships between Y (i.e., output of a model) and the X’s (inputs) must be clearly defined in terms of their transfer functions as well as applicable decision rules and system constraints. The fourth step in the modeling process is collection of process data to aid in development and subsequent analysis of the model. Depending on the model’s objectives, these sources of data include process maps, actual process data related to throughput rates, yields including rework and scrap, machine and direct labor cycle times, downtimes, lot sizes, inventory levels, floor layouts, and other relevant operational data. The specific data that is collected and analyzed by a team should correspond to the questions that must be answered by the simulation model.
Although there are many sophisticated off-the-shelf software packages that can be used to build and analyze simulation models, it will be useful to discuss a simple example to show the basic concepts of more complicated simulation models and software. Figure 5.6 shows the basic steps necessary to conduct a simulation analy- sis. The first step is defining the functional relationships between the model’s out- puts and inputs, represented by the expression Y = f(X). In Step 2, the system’s clock is set to time t = 1, an event is simulated, and the model’s statistics are updated by the software. If the model is not at its terminal time period, the clock is advanced to the next time period, t = 2, and the simulation cycle continues until the terminal
Start Calculate Simulation Model y = f(x) and Database Advance Clock to
the Next Event
Terminate Event? Any Conditional Events? Update Statistics and Variable
States Process Event
Update Model Statistics Stop Yes No Yes No
Using Lean Methods to Design for Process Excellence n 143
time period has been reached. Statistics are collected based on the specific structure and functional form of the simulation model. The functional form of a model will depend on the workflow modeled as well as the probabilities of the event occurrences at each operation within the workflow. We will discuss two simple examples. These include a single operation and a workflow that consists of three sequential steps or operations arranged in series. In Figure 5.7, a single operation is shown in which the value of an independent variable, such as cycle time, has been assigned probability values over its actual observed range of cycle times. A specific cycle time of 12 days will be discussed by way of an example. Using this model we will generate random numbers having a uniform occurrence probability between 0 and 1. These random numbers will be transformed using the cumulative density function (cdf) of cycle times of the actual observed distribution. The cdf has a range between 0 and 1 based on the original probability density function (pdf). The relationship of a cdf to an independent variable can be discrete or continuous. The functional relation- ship between cycle time and its occurrence probability has been discretely defined
Probability Density Function (pdf)
–50 0 50 0.00 0.01 0.02 0.03 0.04
Value of Independent Variable
Pr ob ab ility D ensity F unction 12 Days Random Event Simulation Between 0 and 1 –50 0 50 0.0 0.5 1.0 Cum ul ative D ensity F unction
Cumulative Density Function (cdf)
0.6
12 Days
Value of Independent Variable
in Table 5.5. In effect, any random number in the range of greater than 0.539828 and less than 0.617911 is defined as a discrete cycle time of 12 days. However, using a continuous cdf, we could map a one-to-one relationship between a continuous random variable in the range of 0 to 1 and a continuous range of cycle times. This concept is shown in the lower portion of Figure 5.7. An example, using cycle time as a continuous variable, will be discussed using Figure 5.8 to Figure 5.11.
Figure 5.8 to Figure 5.10 show three common probability distributions used in simulation applications. There are also many others that are used, depending on the distribution characteristics of the system. Once statistical sampling has shown the pattern or distribution of the metric of the workflow under analysis, which in this example is cycle time, the empirical data is fit as appropriate to a standard probability distribution using goodness-of-fit testing methods. Once the empirical data has been fit to a specific probability distribution, the formula for calculating random deviates from the standard distribution is obtained from one or more of the references listed at the end of this chapter. Recall that this formula is derived using the cdf of the probability distribution. Figure 5.11 shows how these concepts are applied in practice using Minitab software. In this example, the total cycle time through the workflow is estimated by adding the simulated cycle times across its three sequential operations. Operation 1 is uniformly distributed with a mini- mum cycle time of 10 seconds and an upper cycle time of 30 seconds. These two parameters specify a unique uniform distribution for operation 1. Operation 2 is distributed, as a normal distribution, having a mean cycle time of 60 seconds and a standard deviation of 10 seconds. Operation 3 is exponentially distributed with
table 5.5 how Simulation Wo
Cumulative Probability 17 0.0312254 0.758036 16 0.0333225 0.725747 15 0.0352065 0.691462 14 0.0368270 0.655422 13 0.0381388 0.617911 12 0.0391043 0.579260 11 0.0396953 0.539828 10 0.0398942 0.500000 9 0.0396953 0.460172
Note: If the random number is less than 0.617911
Using Lean Methods to Design for Process Excellence n 145
a mean cycle time of 90 seconds. The total cycle time, across the workflow, is the summation of the cycle time of each of the three operations. Statistical analysis of the example shows the median cycle time to be 144.37 seconds, and the distribu- tion is highly skewed right.
Admittedly, there may be easier ways to build simulation models using off-the- shelf software. This example was presented to show the underlying logic behind