The variety is too great to be handled in one model that fits all - you need to dig deeper into sub-niches related to what workshop(s) we are dealing with, which resources that are used, the specific machines and their specification and layouts, the logistics system that feeds the production unit(s), the demand structure, periodic maintenance tasks, and many more.

With colleagues in a very large facility we have done this a couple of times, and it is never the same. You need to spend time at the facility in order to learn what is going on - demand, costs, quality assessment, the whole thing. This is not a two-week task - it might be a PhD project, which actually I was the supervisor of - so if I say that this needs a lot of effort you can believe me. :-)

Yes, manufacturing systems are all vary in its size, product variety, automation level, layout, etc, etc. However, I believe, that the best universal methodology to address all these and many other issues is discrete event simulation modeling.

The power of discrete event computer simulation (DES) modeling is its capability of testing multiple scenarios of changing the complex highly variable process without disruption your current process in order to see what steps are the largest waste, where to focus your efforts for improvement, whether the suggested change will really result in the desired improvement and how much.

DES is the most robust and widely applicable modeling and analyzing technique both in manufacturing and in many other industries: healthcare, finance, retail, etc.

DES has become indispensable and demonstrated its power and efficiency for years in addressing such problems as: - Capacity: How many pieces of equipment of various type are required for the manufacturing shop? - Staffing: How many mechanics, workers and other personal are needed for a particular shift in a unit (department) in order to best achieve operational and performance objectives? - Scheduling: What are the optimized staff schedules that help not only in operational efficiency but also take into account staff preferences and convenience?

-What is the optimal equipment maintenance schedule? - Work pieces flow: Is there a bottleneck formed and at what piece of equipment that hamper production throughput goals? - Resource allocation: Is it more efficient to use specialized resources or pooled (interchangeable) resources (say, oven, equipment, and staff)? Does it make economic sense to keep some product lines (or drop them at all)? -- Optimized facilities layout - Design of the facility optimized workflow Also, the Forecasting methodology: How to forecast and play different scenarios with the future product volumes (demand) for short- and long-term planning and scheduling purposes?

Developing a simulation model for the entire facility requires some time and efforts, but such an investment will also be justified as an indispensable tool as a digital twin of the manufacturing process/facility, provided that the model is properly validated and verified.

Please check this paper:

Article Manufacturing systems modelling using system dynamics: Formi...

There are a number of multi-modal simulation environments to address this problem, which may call for some combination of systems dynamics, discrete event simulation, and agent-based modeling. I have used AnyLogic and Simio, which are both nice general purpose environments to build an appropriate simulation. There are also special purpose sims for specific applications - a Google search for your particular problem may point you in the right direction - it depends on how much you need to customize your analysis.

I have been involved in several modelling projects together with industry (cars, trucks, jet aircrafts, water jet engines, ...), and even if we talk about similar products being made or assembled each case is so specialized that there simply is no model to apply for the next exercise.

Michael Patriksson

Simulation modeling is a general methodology. It provides an environment for designing and building models that fit a specific process or manufacturing facility. It means that there is no one unique model that can be used for everything. The next exercise requires building a new model that would fit that another process or facility (not just applying the previously built model for other purposes and process).

It is similar to designing and building a house using the same elements (building blocks - bricks, walls, siding, etc) and data/information. But very different houses (good or poor) can be designed and built using the same elements depending on how good or bad is a designer and the builder.

The basic elements (building blocks) of a simulation model:

•Flow chart of the process, i.e. a diagram that depicts logical flow of a process from its inception to its completion •

•Entities, i.e. items to be processed, e.g. patients, documents, customers, etc. •Activities, i.e. tasks performed on entities, e.g. medical procedures, exams, customer check-in, etc

•Resources, i.e. agents used to perform activities and move entities, e.g. service personnel, equipment, nurses, physicians

•Entity routings that define directions and logical conditions flow for entities

Typical information (data) usually required to populate the model includes:

• Arrival pattern and quantities, e.g. periodic, random, scheduled, daily pattern, etc.

• The time that the entities spend in the activities, i.e.

service time. This is usually not a fixed time but a statistical distribution.

• Capacity of each activity, i.e. the max number of entities that can be processed concurrently in the activity

• The maximum size of input and output queues for the activities

•Resource assignments: their quantity and scheduled shifts

I am definitely not only referring to the simulation parts of the project - which is of course vital to define right -, but much more crucially, in fact, the way that the design structure is parameterized - i.e., how the mathematical optimization of the structure should be represented by the choice of the design variables. I have been involved in several large such projects, and the conclusion, so far, is that rather few parts of a previous design project can be immediately re-used in the next one - the projects are seldom that similar to each other. And then I do speak about the design variables, and not the simulation part of the project, which always vary a great deal between the projects.

Yes, we know from the start what some of the main building blocks need to represent, but it definitely is a rather substantial and crucial sub-project to actually decide the design variables in the best way. Not only may we be forced - or at least asked - to utilize some tools that the industry that we collaborate with will want us to use - which may render a lot of tweaking at the university's lab -, along the way the collaborators may also want to change several things, which costs both time and money - and PhD student anxiety. :-) But we are quite robust, fortunately.

With some partners we have been very lucky to have - and still have - major project collaborations, concretely with the Volvo Group and their sister companies. And we have been lucky as well that some projects have gone live, and are still up and running. :-)

Michael Patriksson

I understand your points. Indeed, it is much easier if the model can be made parametric and re-used. Unfortunately, it is not always possible because sometimes the purpose of the model is playing scenarios with the totally different facility layout, product nomenclature and other factors. Also, sometimes the rigorous mathematical non-linear optimization with minimizing/maximizing multiple criteria is difficult, but it is suffice to playing a few plausible scenarios and picking the one that meets the performance objectives/constraints even if it is not theoretically the best possible (optimal).

As your focus on design variables, I agree that it is crucially important and should be discussed and coordinated with stake-holders and decision-makers who are the intended users of the model project. Defining the design variables is the integral part of the overall model design. The following steps in model design should always be included:

-Problem description

-Questions to be answered by the model

- Performance criteria (these are also very important and usually defined by the stake-holders)

- Decision variables and their constraints

- Design scenarios to be analyzed using the validated and verified model

-Conclusions and recommendation to stake-holders and decision-makes

As an example of one of my multiple projects I attach a short high level presentation that illustrates the above steps "Process Simulation Methodology to Plan for the Facility Renovation of the Surgical Suite at the Children’s Hospital of Wisconsin".

Thanks for the discussion.

Michael Patriksson Alexander Kolker

My understanding from the above discussion:

1. Defining design variables at the first place will set the basis for further modeling and simulation.

2. Design variables can be generic in nature while simulation models cannot, therefore, design variables can be applicable in multiple projects.

3. While DES is the most robust and widely applicable modeling and analyzing technique in manufacturing and other applications, but it's correctness and applicability can only be decided by the choice of design variables.

Kindly correct me if my understanding is wrong and add if I have missed something.

Thanks for the discussion.

Ishaan Kaushal

I would correct points 2 and 3.

2. Once again, defining the design variables is the integral part of the overall model design. Defining these variables depends on the purpose of the model, i.e. what questions do we want to get answered using the model?

Design variables (or decision variables) are the variables that are changed in the process of modeling (within their constraints range) and affect the model's output. The model's output is checked against the performance criteria to see if the specific values of the decision variables result in improvement (or not). That's why these variables are called decision or design ones. These variables can be generic in the same family of models. If the purpose of the model is changed then a new set of decision variables is usually defined.

3. Correctness of the DES and its applicability is not decided only by the choice of the design variables. Correctness of the entire DES model should be verified, and the model should be validated.

Validation is usually defined as evidence that a model has a satisfactory level of accuracy consistent with the intended application of the model.Verification is usually defined as making sure that the model works according to its intended design. It is possible to have a verified but not validated model (say, no bugs and/or no wrong loops in the code) but the model's output does not reflect the real system behavior. Some techniques for model's validation are: animation, comparing model's outcome with some actual historical system outcome (if known), extreme condition test, experts' insight on the model's outcome, replacing variables distributions with some constants to compare with the easy performed manual calculations, and some others.

Some verifying techniques: model's review with another expert, laws of conservation (entities in - entities out=queue buildup or zero for steady state), review some small modules one by one rather than the whole model at once, input data checking, etc (see e.g. Sargent, R., Validation and Verification of Simulation Models, Proceedings of the 1999 Winter Simulation Conference).