Lecture 47, Laboratory Demonstration, Plant Simulation software part 3 of 3, Dr Amandeep Singh

IIT Kanpur July 201
20 followers
2 Video Views·Dec 3, 2023
Good morning, welcome back to the course, I am Dr. Amandeep Singh and I will take the
Plant Simulation Technometrics in this lecture.
I like to talk about the distributions ok, I have just opened this milling process.
The processing time is 1 minute.
If I took uniform distribution, uniform distribution you know that is also known as rectangular
distribution.
So, where we have just this start and stop, we have the minimum and the maximum value
and we know that the value would lie between these two values, we do not have much information.
So, uniform distribution and triangular distribution and beta distribution, these are sometimes
known as the lack of information or lack of knowledge distributions.
Because we do not have much past knowledge, we just have 2 or 3 or 5 or a very few number
of observations and we do not know what a distribution would it follow.
We have just a minimum value and maximum value, we do not know what is happening in between.
So, we just pick this rectangular distribution.
If you remember, if you know the rectangular distribution is just we have A value and B
value that is all.
So, it is showing if we have to put it in a this way stream, start and stop.
So, I will just put start and stop.
Let me say the minimum time is 2 minutes, ok, and the maximum time is 3 minutes.
So, this is there.
Now, I have just put start and stop, I have not put any stream.
What is stream?
If you know about random numbers, stream is the seed of the random numbers.
So, when we talk about the simulation, simulation, what is simulation?
Simulation, we are trying to imitate the reality and what is representing the reality?
What is representing our actual real objects or subjects?
It is the random number, random number which you are working on.
So, where does our random number start from, for this specific, so that is a stream.
So, if instance there are 2 processes, I put the same stream here, same stream means the
random number will start from the same seed and the successive random numbers would be
same.
For instance, if I put the seed value 2 here in one process and seed value 2 here in another
process.
So, the 9th random number that is selected in this process and the 9th random number
that is selected in this process that would be same, 9th, 5th.
Because the seed is same, so the successive random numbers would be same.
So, it is recommended to pick a random seed, not I would say random seed, but at least
a different seed for different processes to have the good simulation process.
So, if we do not put any stream value here, the software would pick a stream value by
itself, ok.
So, I am not picking any, putting any stream value, so I have just put uniform distribution.
So, we know that the minimum time would be 2, ok, or let me say the time varies from
2 minutes 50 seconds to 3 minutes.
We know that this is the minimum, this is the maximum, ok.
First I will talk about triangle distribution.
Triangle distribution is actually sum of two uniform distributions, two uniform distributions
are there.
Two uniform distributions means uniform distribution is just have the smallest value and the largest
value, we do not know what would happen.
But in triangle distribution, we have three values.
We have the smallest value, we have the largest value, but we have the one value that is in
between, but repeating for the maximum number of times, it is repeating for maximum number
of times that the value is mode.
What is mode?
Like mean, median, mode is there.
Mean is the central value, median, mean is actually the average of the values, median
is the central location value, mode is the value that is having maximum frequency, ok.
So, here we can put, if I pick the triangle distribution, it is showing stream C A B.
Here A is the smallest value, B is the largest value, C is the mode value that is repeating
for maximum number of times.
When it is repeating, I am talking about the past data.
In the past data, I have five or six observations.
I know this is the minimum value, this is the maximum value, but there is one value
which is trying to repeat maximum number of times, so I can pick the triangle distribution.
Based upon the past data, if I have a lot of observations in the past, I can pick normal
distribution, ok, if that fits good here.
In normal distribution, it is asking for its parameters, ok.
The normal distribution, I would say normal distribution statistic, it is asking, it
is asking for stream, now it is asking for mu and sigma.
Mu is the mean or average, sigma is a standardization.
Also it is calling for the lower bound and upper bound.
Lower bound is the minimum value and the upper bound is the maximum value within which our
distribution would lie.
So, we should have the knowledge of the distribution.
As I said, the softwares are GIGO, garbage in garbage out.
So, if you put the right distribution and you have the right numbers and be having the
results of the simulation very close to the realistic conditions.
So, this is the work of a systems engineer to design it in a proper way.
So, if I change this time, you can see, ok, I do not apply, ok, because it is showing,
you know, if I do not put it in a proper way, this is stream mu sigma, I am showing the
value of stream it has just picked.
I was showing the value that is 250 and 3.
The value of mean was smaller than sigma.
So, that is why it said, ok, it is showing the negative value, it is not accepting, would
not accept any other format than is required.
So, let me pick this constant time here and it was, just to make you understand properly,
it was 1 minute I will apply.
So, let me see, if I run it for an 8 hour day and try to see the throughput, open, you
try to see throughput, if the throughput per day is 1167, ok, per hour is 48, ok, and
the total throughput is 389 for an 8 hour day, for 8 hours, ok.
Now, if I change the processing times, open, if I change this processing time to let me
take 2 minutes, ok, and I change this processing time also to 2 minutes, in actual conditions
we cannot change the processing time, if the processing that is has to take, it would take.
The milling it has to, if it has to take 2 minutes, unless we change the tools, we change
the machine or we have the CNC machines or we have the advanced machines, we cannot,
but work on more on the processing times.
The processing times are fixed, yes.
In the plant simulation or in the manufacturing simulation here, we can think of working on
the bottlenecks.
The processing times are fixed.
We can think of designing the layout in a way that the material flow is minimum and
that the total time taken is minimum, total time taken would be minimum, that means, the
total throughput would be the larger, ok.
So, if I change this time, now I have increased the time to 2 minutes, now I run the process
and let me try to see the throughput here.
You can see the throughput was larger before, so it has reduced now, because the processing
time is increased, ok, it is 582 pcs per day, 194 pcs in a 8-hour day.
So, also we can have the setup time as well, for instance, a workpiece is to be manufactured,
it will take 2 minutes for processing and 1 minute for setup.
1 minute for setup means, for instance, some milling is happening, if you have milling
process, then what do we do?
We just rotate the tool and remove the material.
So, it is removing the material from one workpiece, this is 2 minute process, after 2 minutes
the machine stops, this workpiece is taken off and a new workpiece is brought in here
that is the raw one and it will start process on this.
Now, this setup, this setup takes 1 minutes here.
So, this is setup time, this is setup time, this is processing time.
If I induce setup time as well and apply, let us see what happens to my throughput now.
Only in one line I have put some setup time, so the throughput is further reduced here,
it was 197 pcs, so it is 194 pcs now, ok, because the setup time is there now, another
1 minute is being taken.
So, total time taken in the milling process in the line 1 is 3 minutes now, ok.
So, this was a brief introduction about the software, the major or the main objects that
we can use.
Now, I will pick some examples, I have some examples for you to show you different layouts
and also the experiment manager will use, will try to see, to see the simulation that
we can do in the software.
So, this is continuous material flow, I will select 2 lane track here, so traffic circle
or segment table, traffic circle pick and open the model, ok.
So, this is the model, it is already running, so I will just run the model, you can see,
ok, this is the trolley that can pick our material from, so this is source part, some
processing is being done here.
I can see what is the time of the processing here, the time is constant 1 minute, so you
can see it is happening, ok, I will make it a little faster, control panel, I will make
it a little faster and then run, ok, this trolley is running, trolley is continuously
running through the tracks, ok, this is actually about 20 times faster, so the processing is
happening, ok, it has happened for the 9 minutes, it is not stopping, so what is the setting,
so there is no end time, so it will continue for the infinite time, so this is a kind of
a just a cell, an O cell, ok, so cancel, so segment table, open model, this is just
showing a track, ok, how the track is built, so let me pick the user interface, ok, dynamic
statistics, display panel or chart, open, so in this model we can see the chart, what
is the setting, there is no end time, I will put the end time here, 08000, apply, ok, let
me try to run this model, so it has run for 1 day, so entity buffer is there, entity machine
is there, this is similar to that, so we can see the charts here, show chart, so for only
one machine it is showing the chart that ok, this is something unplanned, blue colour is
unplanned, then we have blocked, we do not have any failure here, we have this waiting
time for the about 37% of time it has been working, so this is another example, ok.
Then tools and optimisation, ok, the important is experiment manager, I will just pick the
buffer line optimisation, ok, this is line comparison open model, so in this model what
we have? We have 3 lines, in this line there is throughput with no failure, this is throughput
with failure, throughput with failure and buffers, so there is a buff, there are buffers
kept in between, if I enlarge it, this is buffer, ok, this is another buffer, they have
backup buffers in between, this is throughput, we need to see throughput with no failure,
another line is throughput, we have put some failure rates here, ok, if I see the failures,
it has 90% failure, 90% availability, it is 10% failure, in this case there is no, in
the first case there is no failure, ok. So, let me try to run this model, yes, this
model has run for an 8 hour day, so for with no failure the PCs those were produced in
a day were 477, with failure is 347, but because failure is there we have put the buffers in
between that can store some of the material, so it is about 417. So, also it is giving
the percentage of the maximum throughput, maximal throughput is 477, so it is with failure
we have 72% and with failure and but with buffers we would have 87% of the throughput
of the maximum value, ok.
Then I can, then in resources, I can see broker and animation, so shift calendar, worker,
ok, I will show you a worker introduction, ok, worker introduction model. So, this is
the worker introduction model, so in this case you can see the workers are just introduced
into machine and if we run this, let me try to make it a little slower, apply, now run,
you can see the workers who are running it, so I can see the number of workers are 2 here
and work pole is there and the broker is there, broker is trying to distribute the work to
different workers, ok. So, this is one of the example.
Then ok, let me come to the major simulation thing, that is the experiment manager, tools
optimization experiment manager, ok, this is one of the models where we have experiment
manager, this experiment manager can show the simulation, ok. If I run this model or
if let me open this experiment manager first, it is a definition and evaluation. In definition,
we have defined the output value, output value of the workstation portion, start fail portion,
it is working and fail, so what output do we need and what inputs do we have,
input is the route delivery and mean time to repair, ok.
So, if I run this model, so also I can define the experiments that is for number of 100 number
of parts mean time to repair is this much 5, 10, 15, then 5, 10, 15, this is for 100 parts,
this is for 200 parts again. So, these number of experiments would run if I run this experiment
manager. Let me try to run, it has run for 8 hours, ok. Now, let me try to see experiment
manager. Now, I need to see the results. It has shown the results that the mean time to
failure if it is 5, so it is working for this much number of time, it is failed for this much
number of time. So, for the simulation of 100 parts, for the simulation of 100 parts again,
this is for 200 parts. So, it has taken 8 experiments.
So, I can see the report as well here, ok. In the report, it is telling that ok,
there are number of experiments in experiment 1, this happened in experiment 2, this happened,
ok, this is the total failure, this is what blocked time, then grey colour is waiting time,
all these things it is telling in the report. So, in experiments, what has what is it taken?
It has taken different random numbers. In experiment number 1, different numbers are,
random numbers are taken, experiment number 2, different numbers are taken.
What is experiment number 1? It was given here, the inputs that we gave, output values,
sorry output values are here in this result. Experiment number 1 is 100 parts with mean
time to failure as, mean time to repair, mean time to repair as 5 minutes. Experiment number
2 is 100 parts with mean time to repair as 10 minutes, ok. So, it is showing these experiments
here. So, we can have all different kinds of design, then rule, we have not defined
any rule break setting can also be done, this is and like if we go to the detail of the simulation,
these things are possible. So, now, ok, layout optimisation,
factory optimisation, factory layout is there, this is a factory and we have made weird kind
of connections here. So, layout optimisation can be done, for where to keep what machine,
then what would be the overall, what would be the maximum throughput if you do that.
So, this can, these models can be also we can put through these things to find the optimisation
here. So, another model I can pick here is from continuous material flow, tool and track,
distance control, ok, sensors. Sensors, sensors are like if we need to, for instance, if we need
to accelerate at some point or we need to put brakes or we need to put the, this is a light
and like we can do anything, let me open this model, ok, yes, this is an brake and accelerate
model. So, I have defined the method here, watch method, the method here is brake and accelerate.
So, if I try to run this model, ok, first let me see the event simulator, it is n time,
ok, nothing, but just to see whether how it runs, you can see this is brake, after brake it will
just slow down, after accelerate it will accelerate, this is between the speed. So,
at this point brake would apply, it is accelerate now, the brake would apply,
it will take a turn, accelerate, it will accelerate from this point, ok. So, this is a kind of an
entity, this car here, ok, we had just one kind of entity, this is another entity which is in the
form of car, ok, the name is autosorter. So, this was my sensor in continuous material flow.
There are many models in this, ok. So, I will just pick randomly something, some,
ok, tools and optimization, I can pick experiment manager, then I have, I will try to explain the
experiment manager in more detail, I will pick an example. So, before that, let me pick something,
ok, transfer station, ok, transfer station, I have a demonstration model here.
Transfer station is, this is one cell, ok, this is one cell, this O is one cell, this is another
track here, two-way track. Transfer station, it will just transfer the material from this cell
to this track. Let me try to see the event simulator here. You can see the trolley or
the container came here, ok, the container is coming, some machining is happening here,
some machining is happening at this parallel process, ok. Let us first see what are the
process parameters here. This is a parallel process in which 2 x 2, 4 processes are there,
the time is again the default time 1 minute, ok. So, from source it comes and transfer station is
there. We need not to connect transfer station. Transfer station, when we see the transfer
station, so it will transfer, it will load, it will just load the process, it is connected,
connect path from the parallel process, we just need to put the name of the process,
from which process, it will pick the path from parallel process, ok.
Then it will target or the transfer the parts to line, ok. So, there is no connector required
in between. We need to just mention the predecessor and the successor processes here. So,
sensor position is this one, so all these. So, we can select load, unload, reload, move. So,
it is loading the parts. So, all these attributes availability is 100 percent. We can select this,
ok. I would not change anything and let me try to run the simulation. So,
these containers are running. Some processing is happening here. It will take me 1 minute.
So, transfer station is then transferring the part to this container. So, this container is
just coming here and this transfer station which is connecting the part from this line 1 to the
2 way track, ok. If I see its properties, it is from line to the track. The name of this track
is track only, ok. So, again from this, there is another transfer station, ok.
There is source transporter as well. The source transporter that is a source is there that is
trying to transport material from some other frame or some other using some other interface,
it is trying to do that. So, this is one of the objects that can be used. So,
this is the transfer station, ok. Now, slide, ok. So, there are certain examples which are
available for us to see how these things happen. So, I will just open the start page again and
try to open this factory simulation that we had made once. So, this is a factory simulation in
which, ok, let me close these examples, ok. This is a factory simulation in which these
processes are there. This is the process. If I put the name, this is the process,
single process 1 and single process 2, single process 3 and the workers are working here.
The times for the workers are put. The speed of the worker which is defined by International
Labour Organization is 80 meters per second and the times for the different machines are taken
in a way that if it is an automated machine, so the time is actually noted while doing live
experiments. So, these live experiments were conducted. So, this is single process, ok.
This is milling. This is single process. I have not changed the name, but this is milling. This
time is kept constant here and this time is 1 minute and 10 seconds, like that is about 70
seconds it takes to do the machining. Then setup time is about 26 seconds. So,
for this machine, these are actual times which are taken from the experiments,
live experiments are conducted. So, these times for the second machine,
these are the times, the about 2 minutes, that is 1 minute 55 second, 59 seconds for
the processing and for setup, it takes 33 seconds, ok. So, these are all taken constant here. This
is run for an 8 hour day, ok and also an experiment manager, we have defined the
output values. Output values is one is working and fail and throughput of line 1 and throughput
of line 2 we need to see. Like we can just see the throughput here. We can just see the throughput
here in the previous way we did. So, we can just see the throughput of this run that we have done.
It is 534 pieces per day, 178 pieces in an 8 hour day, 178, ok, but if I run my simulation using my
experiment manager, so when I define the experiments, I would define 25 experiments
here, ok, 25 experiments and these processes, processing time is taken actually 59 seconds
and this 59 seconds using random number table or using some random distribution,
normal random distribution maybe. So, we have selected the random times. Those random times
are put here. 59 seconds was the mean and this is the real random times, not 59,
actually 70 seconds is the time, 70 seconds. So, 70, it has 69, this is 69, 1 minute,
9 means 69, this is 70, 71. So, these are the random times, ok. Setup times are also random.
So, these times are put in, time for milling, processing, milling setup, drilling, processing,
drilling setup, grinding, processing, grinding setup. Then after grinding we have inspection,
ok. We have these machines here. Go back. This is my milling machine. This is drilling. This
is grinding. This is inspection 1. This is throughput through line 1, ok. Then worker,
certain pace of the worker, footpath, all those things are defined here. This is the model we
made to, this is an hypothesized factory, ok. So, define observations per experiment. So,
there are 25 experiments, 25 experiments which are conducted using random numbers and per
experiments 50 replication would be taken. That is 50 times one experiment would repeat
and all these 50 times the box plot can be made for these 50 experiments. This is one observation.
One observation means 50 experiments. The second observation is another 50 experiments, ok. So,
it is about a total number of experiments that happen is 25 into 50. So, 25 experiments are
made and 50 observations per experiment. Those are given here. This is 50 observations per
experiment. Then input variables are all these times, milling, drilling, grinding, processing
and setup times inspection 1. In the second line we have a modern manufacturing machine which is
micro wire EDM, electric discharge machining, ok. So, this is a modern manufacturing machine. So,
in this case also the name is put. So, these three machines, inspection 2 is there. So,
the two lines if you can see, the two flow lines. First flow line is the conventional machines,
milling, drilling, grinding. Second flow line is our modern manufacturing source,
micro wire EDM, then inspection and drain. So, we can see the both throughputs used
after conducting these experiments. So, output values are all defined. We need to see throughput
2 and also we need to see the working and the failed percentages, ok.
So, let us start the simulation. So, observation, current experiment is run if you see. A third
experiment is running, fourth is running now, fifth is running. For sixth experiment,
50 observation. For each experiment, 50 observation. For all the 25 experiments,
50 observation will run, ok. So, you can see the time 8 hours, 8 hours, 8 hours it is running.
The multiple runs here, ok, this is how the simulation is conducted using experiment manager.
This is the exact, that is what the, use the word simulation. So, it has generated a report,
total running time for the simulation is 36 seconds, 25 into 50 is 1250. It has run 1250
experiments in 36 seconds, ok. So, this is my report. So, let me take it here. I would not
close it. This is my report. It is saying, ok, this is the simulation. This is the final picture
that we are getting here. This is simulation. Experiment 1, if you would like to see,
it has shown that, ok, single process 1, station 1 and single process 1. It is showing that these
are the time for which it is working. These are time for which it is blocked. It is showing all
these things. Experiment number 2, ok, all these things it is showing. So, it is showing, let me
try to see the statistics here. So, mean value, standard deviation, minimum and maximum for
experiment 1, ok, for all the experiments. So, this is the P and T test. We would not move to
that. These are the tests those are conducted and this is the chart for that, not chart,
this is table for that. This is the failed time, working time and also it will show the throughput,
because we asked for the throughput. So, this is the throughput for the 50 observations in
experiment number 1. Experiment number 1 means 50 observations. There are 25 experiments,
for each experiment 50 observations. So, experiment number 1 has 50 observations.
These 50 observations have this mean 177.4, ok. These 50 observations have this standard deviation.
It has this minimum value and this maximum value, left and right bound is given. So,
for all these 25 experiments it is taken, the overall mean would be also one throughput that
we also see, ok, for throughput of line 1 and throughput of line 2 we can also see here,
ok. These are also again for 50 observation 25 experiments. So, let me try to see, ok,
this is the model, this is overview, this is values of experiments. Let me try to see the
plots of these. So, these are the output values of interest. So, if you see this,
this is each experiment and 50 observations are given. This is for the throughput 2.
You can see now we have a box plot here. So, now, what does this show? This for instance,
this is my experiment number 5, this is my experiment number 5, this line. In experiment
number 5, there are 50 observation and this is box plot, this is my median and you can say there
is a whisker and the third quartile is quite lower and we do not have the first quartile.
First quartile is kind of coinciding with my quartile 2, ok. So, the upper whisker is very
smaller. So, it is showing the box plot for each of the experiments, ok. We can see the
variation. So, you know this is very small variation here in experiment number 11.
In experiment number 16, you can see again there is a big variation on the lower side.
So, you can also see the overall variability is very high, overall variability is very high. Why
it is there? Because the micro EDM or this is actually the behavior or you could say,
behavior of the process only. Micro EDM process, we do not know the kind of experiments which we
did. We did not know that what time would it take. Sometime it took maybe 50, it was actually about
in 100, 600 seconds, it took 10 minutes, sometime it took 15 minutes, sometime it took 20 minutes,
sometime it took like, gate variability was there in the overall process. That is why this big
variability is there, number 1. Number 2 is the number of the throughput is very small. Actually,
it is the final throughput that is obtained, the average throughput was 14 pcs.
Throughput is very small. So, this small throughput is showing very high variation,
like in a day we can even you know we can see that we even have about 9 pcs in a day here,
ok. We have about 15 pcs in a day here. So, the total throughput is very small. So,
that is why this much higher variability is there. So, let me try to see this for throughput line 1.
So, this I have obtained the chart for the throughput of line 1. You can see the throughput
for line, the variability is less, ok. The two reasons for this, number 1,
the processing times were in control because the machines were CNC machines when we conducted the
live experiments. However, the processing time for the inspection was little variable because
inspection was done manually. So, but overall because there were 3 automated machines. So,
the variability is lesser than that we observed in our line 2, ok. This is first reason.
Second reason is this number, the throughput value is 178, the overall average is 178 which
is a big number and like bigger number than that very small number 14. So, the throughput
variability is lesser. However, in individual processes because the simulation is trying to
re-imitate the realistic situation. So, in realistic situation we can have the outliers,
we can sometime have the very long time or very lesser time than it should be normal.
So, this for instance in experiment number 5, you can see again the whisker is falling down.
It all depends upon the random numbers and if I tell you the times were kept constant,
but the seeds were all different. So, depending upon different seeds for these different experiments
and the observations were also very different, these different box plots are obtained, ok.
So, we can infer that, ok, the variability is lesser in this and number of workers we can,
we can keep varying the number of workers, you know, in this experiment the number of workers
were employed were 4, you know, 1, 2, 3 workers in line 1 and 1 worker in line 2. We can also
think of employing 5 workers that is 4 worker in line 1, 1 here, second here, third here which
is not yet employed and fourth here and fifth one here. Then we can see, ok, if we employ another
worker, we can do cost analysis as well. If we employ another worker, what is the
wage of that worker and what is the overall throughput? What is the increase in the throughput?
If the increase in the throughput is lesser than, like the profit that we obtain from the increased
throughput by employing the fourth worker is lesser than the wage that has to be paid to the worker,
so we can just take off the decision, we can just deny this decision. So,
the certain experiments that we can do, certain simulations that we can do,
I think I should stop here. We have discussed enough about the technomatic software for an
amateur. This is a good start, like you have an introduction to how the softwares work and how
is simulation conducted in the software and what are various kinds of layouts and we can even
optimize the layouts using these softwares and there are certain, you know, as I have told you,
there is a great capability of this specific software, but yes, we can keep on using this,
we can try to simulate a factory, we can make a big factory and this is just a manufacturing
process in manufacturing systems. We can also think of taking up to this
manufacturing system, this is developed, this system or the report that we have got,
we can take that report to team center to finally support when actually things happen. So,
we can just put, okay, this is the schedule or the plan, these are things which are actually
happen. So, what is the variation of the actual things which are happening, you know,
first few runs when we actually run this kind of setup. So, we can test those as well using
these software tools. So, I like to stop here and thank you for being in the course. So,
we will meet next time. Thank you.
