I work with the assessment of production system configurations with respect to operational, economic and environmental performances.
A production system configuration is meant to be a facility with specific equipment, layout, numbers of operators, etc.
The assessment involves a comparison of the configuration being studied with other alternative configurations (for instance, a case "yes, new equipment" VS "no new equipment, status quo"). The comparison is being done through discrete event simulation.
Now, my doubt is on how to make sure that the evaluation compare the two cases fairly.
Beyond the fact that the two systems get supplied with the same input distribution of pieces to process, I wonder if I have to make sure that the two systems are evaluated and compared with the same level of performance operations wise (e.g., same yield, same efficiency).
To me, this is to make sure that both systems' operations are managed by the same people having the same expectations in terms of efficacy of operations management.
An example of this approach is to optimizing both systems firstly and then compare the resulting performances.
What are your thoughts on that?
Every idea and source is most welcome. I will particularly appreciate previous experiences in discrete event simulation of production systems and considerations about design of experiment in the context of production systems and operations.
Howdy Miss Beretta,
In the field of simulation, a discrete-event simulation (DES), models the operation of a system as a discrete sequence of events in time. Each event occurs at a particular instant in time and marks a change of state in the system. Between consecutive events, no change in the system is assumed to occur, hence, the simulation can jump in time from one event to the next or the other.
This is in contrast with continuous simulation systems in which the simulation continuously tracks the system dynamics over time. Instead of being event-based, this is called an activity-based simulation; time is broken up into infinitesimal (small) time slices and soing so, system state is updated according to the set of activities happening in each time slice. This results in a continuous simulation system, though each time slice represents a discontinutiy. SInce discrete-event simulations do not (have to) simulate every time slice, they can run faster than the corresponding continuous simulation systems. But hthey lack the dynamics!
Another alternative to event-based simulation is process-based simulation. In this approach, each activity in a system corresponds to a separate process, where a process is typically simulated by a thread in the simulation program. In this case, the discrete events, which are generated by threads, cause other threads to sleep ro be awake and hence update the system state.
A more recent method a three-phased approach to thediscrete event simulation. In this approach, the first phase consist of a jump to the next chronological event. The second phase is the execution of all events that unconditionally occur at that time (called B-events). The third phase is to execute all events that conditionally occur at that time (called C-events). The three phase approach is a refinement of the event-based approach in which simultaneous events are not cahotic but ordered. This makes the most efficient use of computer resources. The three-phase approach is used by a number of commercial simulation software packages, but from the user's point of view, the specifics of the underlying simulation algiotithms are generally hidden (for commercial reasons)
Anyhow Miss Beretta, Here is a list of DES software.
Adevs C++ Library February 20, 2014 (v2.8.1). BSD A C++ library for building discrete event simulations. Adevs is based on the Discrete Event System Specification (DEVS) and Dynamic DEVS modeling formalisms; it supports parallel discrete event simulation and a runtime system for OpenModelica.
CPN Tools BETA Software September 1, 2013 (v4.0.0) GPLv2. A tool to analyse logistics/queuing models in all types of applications.
DESMO-J Java Library March 26, 2014 (v2.4.1). Apache 2.0 A framework for discrete-event simulation in Java, supporting hybrid event/process models and providing animation in 2D and 3D.
Facsimile Scala Library April 29, 2009 LGPLv3. A free, open-source discrete-event simulation/emulation library.
Galatea Java November 4, 2013 (V3.0r6) GPLv2. An agent-based simulation platform.
JaamSim [6] Java Software April 8, 2015 (v2015-18). GPLv3 JaamSim is a free and open source discrete-event simulation software which includes a drag-and-drop user interface, interactive 3D graphics, input and output processing, and model development tools and editors.
MASON Java Library November 29, 2013 (v17) AFLv3 .A fast discrete-event multi-agent simulation library core in Java, designed to be the foundation for large custom-purpose Java simulations.
PowerDEVS C++ Software September 18, 2013 (v933). AFL, GPLv2 An integrated tool for hybrid systems modeling and simulation based on the DEVS formalism.
Ptolemy II Java Library October 28, 2010 (v8.0.1) BSD. An open-source software framework supporting experimentation with actor-oriented design.
SIM.JS JavaScript Library January 16, 2012 (v0.26). LGPL SIM.JS is a general-purpose Discrete Event Simulation library written entirely in JavaScript. Runs in browser and a GUI-based modeling tool is supported.
SimPy Python Library April 7, 2014 (v3.0.4) MIT An open source process-oriented discrete event simulation package implemented in Python. It is based on Simula concepts, but goes significantly beyond Simula in its synchronization constructs.
SystemC C++ Library March 23, 2014 (v2.3.1). SystemC an Open Source License A set of C++ classes and macros which provide an event-driven simulation kernel in C++.
Tortuga Java Library September 1, 2008. LGPLv2.1. An open source software framework for discrete-event simulation in Java.
References:
"Adevs - Sourceforge release list"
"CPN Tools - Whats New?"
"DESMO-J version history"
"Facsimile News"
"Galatea - Sourceforge release list"
JaamSim homepage
JaamSim downloads
"MASON - Homepage"
"PowerDEVS - Sourceforge release list"
"Ptolomy II Homepage"
"SIM.JS - code.google.com changes"
"SimPy - Bitbucket tag list"
"accellera"
"Tortuga - Download list"
Have fun!
Frank
Dear llaria,
According to my industrial experiences, I'd like to suggest you to calculate and compare the average cost which contains average fixed and average variable cost regardless who is the operator. Average cost is the most practical measurement for performance evaluation. It tells you the efficiency of both the operator and the system operated by the operator since it reflects the marginal cost based upon the marginal cost.
Hello illaria,
I'm struggling to understand your question.
What is the purpose of the simulation you're doing? Is it to compare an existing condition to possible alternative ones (for an existing site)? Or to compare different options (for a new site)? If either is the case, then the systems will inherently have different efficiencies. Which is precisely what you want to find out. If they all had the same efficiency, then making a decision between them would be as simple as calculating the investment cost for each (just choose the cheapest one).
In my mind, the process should be to create a simulation that matches reality as much as possible (easier said than done), create the different alternatives that you want to compare, and then run them using the same input distribution (as you mention). The efficiency of the system (whatever the efficiency metric is for the issue you're simulating), will be one of the comparison metrics. There will probably also be quality and cost considerations (at least).
One thing to consider carefully as well is the input data. Most systems in the real world will need to cope with different situations (different demand patterns, different processing conditions, different levels of raw materials, differently design products, etc), and it may be possible that some of your configurations are 'better' than others at dealing with some of these conditions. Just because a system is better than another one under some conditions, does not mean that it is always better. You may also want to do sensitivity analysis around process parameters where there's some uncertainty.
Cheers!
Eduardo
Hi -
Regarding the comparison - as usual the answer is: it depends.
In particular it depends on what aspect of the system you are trying to evaluate. If the intent is to understand performance differences then of course you would leave these differences in.
In our studies of real systems though we are often faced with production rate variation caused by factors we are not interested in (supplier side problems or particular machine downtimes). In particular when trying to understand physical workload on system operators the current rate of production had a huge impact and then, to get a "clean" comparison we would normalise or adjust our variables to eliminate the effect of production rate differences. At times using DES has helped us overcome this (see our car recycling paper led by Kazmierczak).
Its all about isolating the particular aspect you are trying to study.
Sometimes I look at the data both ways, which can also provide some insight.
I see you are at Chalmers - you might try talking with Lars Medbo and/or Per Medbo in the Transport and Logistics division. They both have experience thinking about these issues and could perhaps help with your particular situation.
good luck
-P.
I think you are probably looking at this the wrong way round. It is not really a meaningful comparison if you start with assuming the two production systems are working at the same level of efficiency. You need to make the comparisons based on requirements of demand (quantity and variation) and input requirements (materials, labour etc), plus any relevant constraints. You can then use (operational) efficiency as the measure to compare the systems. Of course you still need to decide how to measure "operational efficiency" in the cases you are studying.
Dear all,
big thanks for your precious inputs.
Going back to my question, I understand that the title of it might have been difficult to grasp, even though you still did manage to provide me with the answers that I needed. Therefore I thank you again for this.
For "same operational efficiency" (which are not the proper words to describe the follow) I meant having the two systems being designed and run at their own best possible ways. In making this assumption, I forgot that that one would be the best possible way according to a certain condition, but as input demand and operational condition vary over time, modeling and comparing the two systems according to that condition would be unrealistic and non-robust.
Following my little background in simulation which got now strengthen by your answers, I truly understand now that the two systems have to be modeled on the basis of to the same input demand and requirements, compared according to the output performances (efficiency might be one of these) then studied further according to the factors we are interested to assess or isolate.
I hope I will have the opportunity to have further discussions and benefit from your inputs in the future.
Best regards
Ilaria