Over the years I’ve used simulation to address many different types of problems.   There are traditional questions about throughput, manpower, utilization, and capital spending.  For these questions a model is built to represent the operation of the system and variables are changed.  Once we feel the model fairly represents the system (verification and validation), we try to maximize some sort of objective function that summarizes the questions and concerns about the system.  We usually do this through some battery or experiments or maybe even a structured design of experiments process.  This is the foundation of our industry.

The concepts behind simulation have been applied in many other ways.  We have all built models for, shall we say, “non-traditional” purposes.  I’ve talked about how models communicate ideas.  The ability of a model to draw participation from a group is undeniable.  Often models justify themselves strictly because they get a project team on the same page and give them a common frame of reference for a complicated discussion.

So what else?  I have heard people say that if a model does not have a stochastic process that it is not worth doing.  I disagree with that sentiment.  While almost every system in the real world has a stochastic nature, questions about the system can be answered without that level of detail.  I have seen many valuable models that do not include probability distributions of any type.  For example, an automatic paint system for large trucks.  The amount of time required for a robot to paint the truck is different based on the type of truck being built.  A different program is downloaded to the robot based on the truck type.  It is constant for any type of truck but different based on the production schedule fed into the system.  This becomes a scheduling problem with the additional variables of conveyor speed, physical length, flash time for the paint, and cure time in the oven.  Through heuristics, we can evaluate the rules for how to process jobs effectively.  Is it better to batch like trucks through the system or find pairs of vehicle types that make the process flow with a higher utilization?

What about other applications?  Our latest application of discrete event technology is PLC emulation.  We create a virtual model of the hard automation on a factory floor and let the control for this system be driven from an outside source.  This outside source is usually a PLC but in reality it could be any outside algorithm.  The model in this case simply responds to the commands from the outside intelligence and provides feedback.  If the PLC tells the model to turn on a conveyor motor, the conveyor turns on.  If a part on that conveyor trips a photoeye, that event is reported back to the PLC so it can decide what the appropriate reponse is.  These models by definition are deterministic.  We are trying to validate that the logic in the PLC will perform appropriately when used in the real world.  Essentially, the PLC program is the variable that we are changing and the model represents the part of the experiment that stays constant.  This application has an incredible payback because it allows code to be validated before the real world is either created or interrupted.  It says valuable factory time, finds errors earlier in the process, and allows for testing things you woudl never try in the factory for safety reasons.

I have also seen examples of simulations being embedded in many other decision making processes.  Please comment on the post and let me know other applications that you have seen.  This will allow us all to see the many applications of simulation that are right in front of our faces on a daily basis.