Discrete Event Simulation Software: What It Is, How It Works, and How to Choose
Spreadsheets say the line is balanced. The floor says otherwise. Here is the simple idea that explains the gap, and how simulation closes it, in plain language.
A production line that looks perfectly balanced in Excel can still pile up work, starve stations and miss shipments on the real floor. The reason is simple, and once you see it you cannot unsee it.
The short version
- Discrete event simulation (DES) jumps the clock from one event to the next, so a full 8-hour shift runs in seconds.
- Spreadsheets use average cycle times. Real floors have variation, breakdowns, changeovers and blocking, which averages hide.
- Queues do not grow in a straight line. Push a station from 70% to 95% busy and the waiting line can jump from about 2 jobs to 19.
- One simulation run is a guess. Run it about 30 times to trust the answer.
- Modern cloud tools like Calypsion LinkSim run in a browser, with 2D flow, 3D layout and automatic delay tagging.
Why spreadsheet plans fall apart on the floor
Most capacity plans start the same way. You take the average time each station needs, divide the shift by it, and conclude the line is balanced. It is fast, and it feels right. It also quietly assumes every job takes exactly the same time and every machine runs without stopping.
Real floors do not work like that. A spreadsheet leaves out the things that actually decide your output:
- Cycle times wobble. Material, handling and small tweaks make each job a little faster or slower than the average.
- Changeovers cost time. Switching between products breaks the steady flow.
- Machines stop. Short breakdowns starve the stations downstream and block the ones upstream.
- Resources are shared. Operators, technicians and forklifts cannot be in two places at once.
- Buffers fill and empty. A full buffer stops the station before it; an empty one idles the station after it.
The non-obvious part: queues explode
Here is the idea that trips up almost every spreadsheet. As a station gets busier, the line waiting for it does not grow gently. It grows faster and faster. For a simple single-station queue, the math looks like this:
The takeaway: a station that looks "85% loaded" on paper already holds more than double the work-in-progress of one at 70%. Simulation shows this build-up on screen before you commit money on the floor.
So what is discrete event simulation?
Think of a factory as a series of events that each happen at a specific moment: a part arrives, assembly starts, a machine finishes, a breakdown begins. Between those moments, nothing changes.
Discrete event simulation uses that fact. Instead of ticking forward second by second, it jumps straight to the next event and skips the quiet gaps. That is why a complex shift with thousands of parts can run and be analysed in seconds.
Every model is built from five simple parts
Entities
The things that flow through: parts, pallets, batches or orders.
Resources
What entities need to be worked on: machines, operators, AGVs, fixtures.
Events
A moment that changes the state: an arrival, a start, a finish, a breakdown.
Queues & buffers
Where entities wait when the resource they need is busy.
Clock & event list
The engine that keeps time and knows which event happens next.
Watch the clock jump
A tiny cell with one assembly station and one packaging station shows the whole trick. Notice how the clock skips the idle minutes:
Job 1 arrives. The station is free, so assembly starts and a "finish" event is booked for 06:05.
Job 2 arrives while the station is busy, so it waits in the buffer. The clock skipped the 3 quiet minutes to get here.
Job 1 finishes and moves to packaging. Job 2 leaves the buffer and starts. A new finish event is booked for 06:10.
In a real model, each time is not fixed. It is drawn from a range of likely values, which is exactly what creates the starving, blocking and lost output you see on the floor.
What factories actually use it for
Teams reach for simulation whenever a change is expensive, risky or hard to undo. Rather than experiment on a live line, they test the idea in software first.
Find the real bottleneck
See which station truly limits output, and how it shifts as the product mix changes.
Test "what if"
Measure the gain from an extra shift, machine or operator before you approve the spend.
Balance the line
Try different layouts and station loads without stopping production.
Size buffers & routing
Find the buffer size and routing rules that stop stations blocking each other.
Plan changeovers
Find the batch sizes and sequence that lose the least time to setups.
Model AGVs & handling
Spot congestion and charging delays for forklifts, AGVs and mobile robots.
Plan maintenance
See how breakdowns ripple across the floor and where to focus prevention.
DES vs agent-based vs system dynamics
Simulation comes in three main flavours. For most shop-floor questions, discrete event is the right starting point. Here is the quick way to tell them apart:
Discrete Event (DES)
Parts moving through stations and shared resources, step by step.
"Where does the line jam if orders rise 15%?"
Agent-Based (ABM)
Many independent "agents" following their own rules in a shared space.
"How do 12 AGVs and 6 forklifts share one narrow aisle?"
System Dynamics (SD)
High-level stocks and flows over time, without tracking single parts.
"How does safety stock swing across plants over 12 months?"
Because factories are naturally made of discrete jobs moving through defined machines, DES is the foundation. When human or vehicle movement drives the result, agent-based behaviour is layered on top.
How to choose the software: a 10-point checklist
Judge a tool on how fast you can build a model, how trustworthy the results are, and what it costs to own. Run any vendor through these ten questions:
- Methods supported. Does it do DES natively, and add agent-based logic when you need it?
- Time to first model. Can you import stations and cycle times from a spreadsheet or CSV, with no code?
- Handles variation. Built-in ranges, shift calendars, breakdowns (MTBF/MTTR) and shared operators?
- Statistical rigour. Does it run many replications and give you a 95% confidence range?
- Automatic delay tagging. Does it label why each job was delayed (breakdown, setup, operator, blocking)?
- 2D and 3D views. Fast flowchart logic, plus a scaled 3D layout for reviews?
- Where it runs. A normal web browser, or a heavy desktop install with a special graphics card and a licence dongle?
- Security & sharing. Cloud collaboration, access controls and safe project sharing?
- Live data. Can it pull real cycle times from your MES or IIoT platform?
- Total cost. Licence model, hardware upkeep and how long training takes.
The five families of tools
Calypsion LinkSim
Running your first simulation in 7 steps
The biggest mistake is building one giant model of the whole plant. A small model that answers a single question is faster and far more trustworthy.
Ask one clear question
Pick a measurable goal and a few metrics to watch (throughput, queue length, utilization).
Too vague: "model the whole plant".Sharp: "Does a second packer lift Line 3 from 100 to 120 units/hour?"Map the flow
List the stations, buffer limits, routing rules and shared resources. A whiteboard sketch is enough.
Gather real data
Collect cycle times as ranges, not single averages, plus setups, breakdowns (MTBF/MTTR) and shifts.
Build the simplest model
Represent the core flow only. Leave side processes out until the basics run correctly.
Validate against reality
Run it on last month's data and compare with actual output. Fix the gaps before testing anything new.
Run it ~30 times
One run is luck. Many runs give an average with a 95% confidence range you can act on.
Change one thing, then deploy
Test a single variable, compare to the baseline, pick the winner, roll it out and confirm the real gain.
Simulation without the heavy desktop
Calypsion LinkSim™ is a cloud-native discrete event simulation tool built for plant teams and industrial engineers. It removes the usual blockers: no heavy install, no licence dongle, no code to maintain, and no mystery in the output.
Runs in your browser
Build, run and share models with no local install and no special graphics card.
2D flow and 3D layout
Draw the logic in 2D, then switch to a 3D view with LinkTwin without rebuilding.
Automatic delay tagging
Every delayed job is labelled: breakdown, setup, operator or upstream blocking.
Replications built in
Runs repeat automatically and report results with a 95% confidence range.
Live plant data
Pulls real cycle times and downtime from LinkVision straight into the model.
Frequently asked questions
What is discrete event simulation software?
It is a digital tool that models a process as a sequence of events, such as job arrivals, machine starts, breakdowns and completions. It jumps the clock straight from one event to the next, so you can test layout, capacity and scheduling changes on a computer before touching the real plant.
What is it used for in manufacturing?
Finding bottlenecks, balancing lines, testing capacity before buying machines, sizing buffers, planning changeovers, routing AGVs and planning maintenance. The goal is to prove the return before spending the money.
Do I need programming skills?
Not with modern tools. Older platforms need custom code, but cloud tools like Calypsion LinkSim let you build models visually and import data from a spreadsheet or CSV file.
How accurate is it?
It depends on the input data and validation. Feed it real cycle-time ranges, breakdown logs and shift calendars, check it against last month's actual output, and a model predicts throughput and queues reliably.
What data do I need to build a model?
The essentials are:
- Cycle times as ranges, not single averages
- Operator assignments per station
- Shift schedules and breaks
- Changeover and setup times
- Breakdown frequency (MTBF) and repair time (MTTR)
- Buffer limits and routing rules
Is there free or open-source software?
Yes. There are free and open-source options, often Python-based. They are capable but need programming skills, lack automatic delay tagging and 3D views, and offer no commercial support.
How long does a model take to build?
For one production line, a baseline model can be built and validated in one to three days using spreadsheet imports. Collecting the data usually takes longer than building the model.
Stop guessing. See your line run in software.
Model your real bottleneck with Calypsion engineers, or try LinkSim's 2D and 3D simulation live in your browser.