Automizing Factory Optimization
Plant Simulation is a computer application developed by Siemens Industry Software for modeling, simulating, analyzing, visualizing, and optimizing production systems and processes, the flow of materials, and logistic operations. Building Plant Simulation models is a bit like playing Sim City, building a model of the system you want to optimize, programming special behavior every now and then. Optimizing the built model can be a tedious task, that requires significant expertise, especially if system behavior depends on multiple input parameters that interact with each other in a very specific way. Therefore optimization can take a significant amount of time in a project and sometimes the best results are not found due to project time constraints.
Design space exploration and optimization software takes modeling and simulation to the next level by allowing users to determine appropriate values of variables that yield product designs that result in exceptional performance. Siemens Heeds was actually designed to support CAE and CFD simulations but can be utilized with Plant Simulation as well.
Design Space Exploration and Plant Simulation
The Plant Simulation model in the following video was built to represent a manual process manufacturing environment. Performance is depending on times in work order descriptions per order and the order mix that is applied to the environment. Process time on the highlighted station is very heterogenic and might result in multiple parallel stations. The simulation model is supposed to solve the questions of how many workers and parallel stations are needed to make the overall throughput performant. The standard approach to explore these interdependencies would be, to check different worker and station numbers against each other to find suitable setups. Plant Simulation offers Design of Experiments modules to check a multitude of combinations of different stations and worker number values. The result in Plant Simulation shows how the two variables impact system performance. Since Plant Simulation is primarily a modeling tool the result visualization is a bit hard to understand and explain, especially for non-simulation experts.
The advanced Design Space Exploration capabilities of HEEDS allow for a more intuitive view. The 3D graph visualizes how the numbers of workers and stations correlate to throughput. It is visible that using only one station will cap more throughput quickly, using two stations leverages more workers very well and that a third station only adds an unreasonable amount of extra throughput. The 3D radial view also supports this result, since good results can be reached with only two stations.
This way, the functionalities employed by HEEDS Design Space Exploration help engineers to understand overall model behavior, to find sweet spots in the planning landscape, and to explain results and a decision two third parties (like decision-makers).
Design Space Optimization and Plant Simulation
There is even more synergie between Plant Simulation and HEEDS. The following Plant Simulation model represents a dairy factory that processes raw milk into finished milk products. Model behavior is dependant on raw milk intake and production programs that can be impacted by seasonal demand for finished goods. Major levers for performance are the different tank and mixer volumes. Changes to one single volume value affect the behavior of all other tanks and mixers in the model.
The historical approach would be to execute a number of experiments to find a good synergy between all tank and mixer sizes, production program,s and raw milk intake. The quality of the result mainly depends on the planner's experience to find that synergy manually.
Solving the model with standard Design of Experiments, exploring alls possible options, would result in a very long optimization time (5 billion years on a current professional computer) since so many different variant options exist.
SHERPA will guide you
The HEEDS SHERPA optimization environment will learn how different tank and mixer sizes will affect the performance of the simulation model by executing simulation runs in the Plant Simulation model automatically. It will change different variables based on mathematical optimization algorithms, mimicking what a real planner would do- just with the speed of modern computers. Over time SHERPA will learn how to use different variable settings with each other to create an optimized variant set.
Planners can also branch off into Design Space Exploration at any time during the optimization process.
Final thoughts
Utilizing Plant Simulation with HEEDS will allow you to mimic the manual optimization process with proven mathematical optimization algorithms at modern computer speeds not being dependant on human (at human speed) interaction between simulation runs.
The combination of exploration and optimization features in HEEDS allows optimization engineers to branch off from automized optimization at any time- adding human creativity and experience at critical optimization areas and not generally.
The combination of automized and human-driven optimization can deliver better results even in large models with general parameter setups.
