How to Use Computer Simulation to Model Your Success

Computer simulation modeling and analysis is quickly becoming the key to success for any company that is considering:

• Automating their process and/or MHE equipment
• Optimizing an existing automation operation
• Upgrading their automation capabilities and technology

When done right, computer simulation modeling provides a virtual environment in which design ideas can be tested, discretely analyzed, iterated, refined, and proven out — all before a single piece of equipment is purchased or bolted to the warehouse floor.

But to yield a reliable analysis, computer simulation modeling demands a good deal of knowledge and skill. And even when the same methodology is used by two different companies, the results you get can vary greatly depending upon their modeler’s expertise.

So it’s important to educate yourself when considering this critical service.

Here are a few steps that will help.

Step 1: Know What You’re Getting

Simply put, some intralogistics providers offer modeling and simulation and some don’t. So start there.

When you do find providers that offer it, pay close attention to the application, approach, and analysis offered as it can vary in notable ways that make comparison difficult:

• On one end of the spectrum, there are providers that use simulation as a perfunctory means of ensuring that the system they sold the customer will actually meet the throughput levels promised by their design team. For these providers, simulations are primarily a validation tool, designed to save them time in the commissioning process, but they will also likely use it to ensure buy-in from the customer prior to the system being built.

• On the other end of the spectrum, there are providers that use simulation to provide the best possible solution for accommodating their client’s unique business needs. For these providers, computer simulation modeling and analysis is the centerpiece of their sales cycle and an essential tool for not just developing the system design, but for also developing the algorithms that will be used to govern the system.

So it’s important to learn as much as you can about what is being offered and how it is being used.

In the past, a good deal of the modeling for material handling / fulfillment operations was done mathematically through spreadsheets. Those models were designed around averages and used static inputs, which were not reflective of real-world conditions. That meant simulations often produced conclusions that either led to under designed, underperforming systems or over designed systems that cost more than they should.

Today, a methodology known as discrete event simulation is widely used. The term “discrete event” refers to the ability these simulations give modelers to analyze whatever “events” they build into the model.

If your provider isn’t using this methodology, you should shop around.

Discrete event simulation differs from other forms of modeling and simulation in the automation industry in a few critical ways:

1. There is an element of randomness to the simulation that better reflects real-world conditions.
2. There is an element of time in the simulation that enables far more accurate design testing and proof.
3. The simulation can provide a visual representation (animation) of a system design or process along with graphs and tabular data to aid in the analysis.

Randomness ensures simulations are able to be run under varying operational conditions. This eliminates the testing of static values and thereby produces more realistic outcomes. The goal behind iterative simulation testing is to produce consistent results under varying or randomly unique operational conditions. Introducing the element of randomness into a simulation makes that possible.

The incorporation of time into the simulation is a game changer in that it allows modelers to see how events would actually take place over time. Instead of having to work from averages or ranges (e.g. 3 totes arrive at a goods-to-person station every minute), discrete event simulations log the times those totes actually arrive at the station. This is important, because when those totes arrive will determine how quickly/efficiently they are processed. Huge swings in productivity can be seen between a scenario in which totes arrive consistently every 20 seconds versus one in which totes arrive in clumps or with long gaps in between. So knowing when an event occurs is far more useful than knowing the range or average of how often an event might occur.

And, finally, the interplay of model elements can best be understood when they are visually presented in the form of animation. Even rudimentary animation immediately highlights potential problems that can be analyzed in ways that lead to a more comprehensive understanding of the interplay of all elements in a model.

Step 2: Understand the Differentiators

Discrete event simulations empower you with the most accurate understanding of how a design for a fulfillment automation system (or enhancement to a system) will stand up to real world production demands — but only if your model is programmed in a way that nets reliable results.

A computer model is a product of the modeler or engineer who programs it. And when it comes to programming models, experience matters — as does artistry.

Experienced modelers build models based on the questions they want answered in the simulations they run. By building specific events/details into their models, experienced modelers know they will get the answers they need. Put too many details into a model and it becomes cumbersome and often won’t answer any question well. Add too few details and the model is likely to deliver overly optimistic results.

So experience matters in knowing which events/details to include.

Good models not only include critical details about the equipment systems being modeled, but the environment in which the system will operate as well as the business rules that will govern the operation.

For an accurate picture of how a fulfillment automation system will perform in a real production environment, a model should include critical details like:

• Real-world client data pertaining to inventory SKUs, SKU velocity, carton sizing/weight, etc.
• Real-world client order data (volumes, peaks, trends, etc.), including future-state order data achieved through data synthesis.
• The business requirements particular to each client’s fulfillment operation.
• The underlying algorithms that will be used to run the system.

And don’t forget the input variables!

To provide useful analyses, simulations need to reflect the variables that emulate real production conditions. Models built using static values or set rates will not paint an accurate picture of how a system will perform in fluctuating operational conditions.

So it’s equally important to include input variables that pertain to changes in production demand such as the:

• Number of lines per order
• Number of pieces per line
• Time it takes for a worker to perform an activity
• Number of Automated Mobile Robots (AMRs) used in a system / per workstation
• Speed, acceleration, deceleration, rotation, and spacing between AMR pick-up and drop-off
• Number of work stations in use at any moment
• Percentage of work going to particular stations

If the right details, events, and input variables aren’t programmed into a model, a computer simulation is really not much better than a spreadsheet simulation that uses averages and static variables.

Step 3: Ask Twice About the Algorithms

A critical step in ensuring that a simulation will yield an accurate analysis is to incorporate the algorithms for a fulfillment solution into the simulation used to test it. But experienced modelers will do more than just test the algorithms that will run a system.

Experienced modelers also use simulations to develop algorithms.

Running iterative simulations against a model using varying input parameters exposes the break points in any system design. Experienced modelers use this information to enhance or hone the algorithms that will be used to run a system.

By building measures into algorithms that enable a fulfillment solution to adapt on the fly to the variables that would cause most systems to fail, modelers can not only create more robust algorithms that are adaptable to wider swings in production demand, they can even create algorithms that can accommodate future changes in business models.

And by creating more robust algorithms, modelers also create leaner, smarter, and more efficient fulfillment system designs that better utilize resources and require less investment capital.

So ask about the algorithms — how they’re developed, and how they are tested.

Step 4: Watch Carefully

Discrete event simulations can produce animations that are useful in ways that go beyond the graphs and tabular data that are also generated by the simulation. A good animation will:

1. Verify that the elements of the model are programmed correctly by demonstrating the expected system behavior. (e.g. High-speed conveyor merges and sorters will work, AMRs will follow expected paths, etc.).
2. Spotlight the impact upstream and downstream systems might have upon one other. (i.e. A transient queue in one area might lead to gridlock in an upstream area.)
3. Pinpoint bottlenecks, grid lock, and sub-par performance areas without requiring the modeler to create (program) separate graphs for every single merge and divert.

Careful observation of the animation will enable a broader understanding of the system as a whole.

What to Expect

The real world example shown here will give you an idea of what computer simulation modeling looks like when provided by an experienced modeler.

This example was created at MSI Automate for a goods-to-person pallet building operation that used sequential-case-picking to pallets to fulfill orders destined for store locations. Pallet AMRs (also referred to as Pallet Ants) are to be used in place of traditional forklifts.

This simulation enables analysis of the following processes:

• Pallets are replenished then put-away into reserve.
• Pallets containing inventory required for order fulfillment are picked-up by pallet AMRs from reserve and moved to a staging area.
• Inventory pallets are moved from the staging area to goods-to-person (GTP) stations where cases are picked for order fulfillment and assembled onto order pallets that are then delivered to stores.

Variable inputs were factored into this simulation to enable iterative testing and make the results better reflect those from an actual production environment. They included:

• The total number of pallets to be built in the simulation. (Testing for variations in order demand.)
• The number of workers available to build order pallets. (Testing for variations in workforce availability and worker break times.)
• The percentage of work each GTP station receives from a particular reserve aisle as well as the percentage of work (i.e. inventory) coming from each bay within that aisle. (Testing for variations in inventory availability and potential slowdowns around high demand inventory items.)
• Number of lines per order pallet. (Testing for variations in how many different SKUs are required to complete an order and how that effects productivity.)
• Number of pieces per line. (Testing for variations in how many different cases are required to complete an order.)
• The time it takes each worker to pick a case from an inventory pallet and place it onto an order pallet. (Testing for variations in worker productivity.)
• The acceleration, deceleration, and rotation speed of the pallet AMRs and time required for pick-up and drop-off of pallets. (Testing for variations in traffic conditions and navigational challenges.)
• The spacing between pallet AMRs. (Testing how variations in operational congestion affect efficiency of movement.)
• The number of AMRs used in the operation. (Testing for variations in efficiency and optimal resource utilization.)
• The number of AMRs used per workstation. (Testing for variations in resource demand.)
• Time required for inventory pallets to be replenished. (Testing for variations and bottlenecks in inventory availability.)

The results from simulations run against this model revealed critical information our designers could use to better understand how our concept would withstand real world production challenges. Each simulation revealed the following:

• How long it took to complete an operation.
• How long it took to complete each order pallet and how many pallets could be completed in that operation.
• How many replenishments were performed.
• How busy the workforce was in that operation.
• How many pallets were staged and sequenced and how long it took for each.
• How many trips the pallet AMRs took and the trip times for both order pallet and replenishment transport.

By running iterative simulations using wide swings in variable inputs, MSI Automate modelers are able to understand how designs will perform in a multitude of scenarios and then adjust our designs and algorithms to better accommodate those scenarios.

Model Your Success

The process of computer simulation modeling and analysis not only enables in-depth learning in a risk-free virtual environment, it also enables a much smoother, more reliable go-live process, while providing our clients with answers to critical questions that they can use to make better informed decisions.

When clients are able to visualize and analyze how a change in automation will affect their fulfillment operation, they are better able to:

• Understand how specific changes in process will enhance their operation.
• Gain critical buy-in from key decision influencers.
• Ensure designs will perform as promised prior to committing to any purchase.
• Justify the changes that will provide them the strategic market edge they need in today’s competitive market.

To find out if computer simulation modeling could be of benefit to your operation set-up a free consultation with our team.

SCHEDULE A FREE CONSULTATION

Walter High is VP Marketing at MSI Automate, where he has worked since 2012.