I'm a beginner and trying to learn simulation modeling. My aims for simulation includes;

A firm receives 5000 orders per week. 120 products are manufactured in an hour. The machine runs for 4 shifts per day and each shift is of 2 hours. Each shift incurs a breakdown of 30-50 minute random. I'm trying to simulate this process and analyze the total product produced and its throughput.

May somebody suggest me learning resources for this purpose and whether we can simulate it or not? All help, suggestions, and feedback will be deeply appreciated

  • 2
    $\begingroup$ simulate it with pencil and paper $\endgroup$
    – jsotola
    Commented Apr 20, 2023 at 17:33
  • $\begingroup$ I would set this up in Excel, with the Solver. As for the random breakdown I would use randbetween() for the duration (30 to 50) and a randbetween() binary (1 or 0) so it happens or not. $\endgroup$
    – Solar Mike
    Commented Apr 20, 2023 at 18:05
  • $\begingroup$ it is unclear if the breakdown always happens at every shift $\endgroup$
    – jsotola
    Commented Apr 21, 2023 at 1:27
  • 1
    $\begingroup$ Seams like a relatively simple math problem. Please try to solve it, and post your attempt, then we will be able to help you more. If you need to, simplify it first by assuming a breakdown of exactly 40 minutes every shift. $\endgroup$
    – Drew
    Commented Apr 21, 2023 at 2:12

1 Answer 1


A single machine with no input constraints can be modeled as an average.

First of all, if you're going to model manufacturing, model shifts that make sense to people in manufacturing. Shifts in factories are 8 hrs, 12 hrs, or between the two if the plant doesn't run 24 hours a day. I've never even heard of a 2 hour shift. Below 8 hrs, I've seen 6 hour watches in the Navy, where people aren't paid by the hour and live on the job. So say the machine runs 8 hrs a day with 4 40 minute stoppages. (First thing you would do at this plant is fire the maintenance manager.)

So 120 units/hr times (8 - 160/60) hrs = 5 1/3 * 120 = 640 units. You can do this math in your head. There's nothing to model here, you have a production rate times a run time.

If you had multiple machines you would ensure that each had a buffer that insured that when the previous machine stopped that it would have material ready. You would do this based on statistics of downtime; in your example we'd need 50 minutes of stock. You would then determine which machine was the constraint and run it constantly, running the others only enough to maintain the buffer stocks.

If you want to understand factory production you should read:

  • Today and Tomorrow by Henry Ford
  • Toyota Production System: Beyond Large-Scale Production by Taiichi Ohno.
  • The Goal by Eli Goldratt

[if you look at Today and Tomorrow on Amazon, these 3 books show up as "Frequently Bought together"]

The next books to have are:

  • A Revolution in Manufacturing: The SMED System by Shigeo Shingo (other Toyota Production System books are great as well esp those by Shingo)
  • Introduction to TPM: Total Productive Maintenance by Seiichi Nakajima
  • Out of the Crisis by W Edward Deming (quality stuff is deep, this is a start)
  • Lean Six Sigma by John Wellwood (there are a lot of choices here)

There are thousands of books digging in to the topics above, I've given you the seminal works.

  • $\begingroup$ Thank you so much sir. $\endgroup$
    – Satya
    Commented Apr 21, 2023 at 17:35
  • $\begingroup$ So basically here the downtime are greatly due to Quality defects in parts, mostly due to process parameter and technical reasons. management tool like TPM, OEE are already applied. so downtime is not due to management fault. $\endgroup$
    – Satya
    Commented Apr 21, 2023 at 17:46
  • 1
    $\begingroup$ Basically, I have time series data that contains cycle time, the number of products manufactured in an hour, downtime incurred, and issues causing downtime. My idea is to predict the next downtime and its cause. And simulate this entire process so that it acts as a digital twin. I want to do predictive analytics and anticipate the total products that can be manufactured if the production is running for say x days. so is simulation good for this? $\endgroup$
    – Satya
    Commented Apr 21, 2023 at 17:47
  • $\begingroup$ @Satya, you can't predict when a machine will break, it's probability. You can assume that past performance will continue then the cause is most like the most likely previous cause. What you should do is eliminate the causes. Ishikawa & Pareto diagrams would be my first step. You don't need predictive analytics and modeling to say it runs 77% of rate, and the planner doesn't care why it stops, only that you can meet the 77% rate. put your energy into making it run 95%. $\endgroup$
    – Tiger Guy
    Commented Apr 22, 2023 at 22:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.