Industrial Event Takt Times (production cycle times) and Probability

Tags:
1. Jul 20, 2015

SSGD

Background:

I am a Mechanical Engineer working as an Industrial Engineer. I have collected some data that is the amount of time that an event took to complete. I first assumed it would be normally distributed, but after plotting a histogram and a normal distribution with the data, I doubt that the set follows a normal distribution. So lets do a little thought experiment.

What is the probability of the event taking from -2 to 0 second? I would say it would be zero.
What is the probability of the event taking taking from negative infinity to zero seconds? I would say it would be zero.

So I think it is save to say that the domain of the distribution is from 0 to infinity.

Question:

What are some models tailored for time distribution? I would say a time distribution should have a domain that goes from 0 to infinity.

2. Jul 20, 2015

Dr. Courtney

There are several approaches to this problem, but when the mean is more than a couple standard deviations above zero, there are occasions when the normal distribution still works well even though the values really can only be positive.

3. Jul 21, 2015

SSGD

Yeah the mean and the most probable time are close by looking at the histogram vs the normal distribution, but still about a second or so, which is actually a big issue for this study. We perform a lot of these evenst which last about 8 seconds. So 1 second is about 15% of the total time. I am about 2 to 3 STD from zero. I should plot it and post it. It starts off very close to zero for about 6 sec. Then sharp peak about 7 seconds and a slower decline to zero.

4. Jul 21, 2015