Plant simulation: the way forward to the digital twin in smart manufacturing
Written by Twan Huijbers and Orkhan Allahverdiyev— Feb 22, 2023
Manufacturing environments are turbulent environments. They must cope with high customer requirements, raw material uncertainty, staff shortages, fluctuating customer demand and complex planning strategies. A simulation of a factory plant can help a company to get a grip on their plants KPI’s.
Using a 5-step approach for preparing, setting up and using simulation one can quickly predict or optimize a new or an existing manufacturing process environment and evaluate suggested changes assess alternative designs.
Classifying, comparing and discussing different simulation techniques gives the best choice for the job at hand. At the end of this paper two examples of simulation implementations show what choices have been made for those specific situations.
A framework for simulation
In Figure 1, Robinson (2008) gives a clear framework of how a simulation is built. In this framework, there are 5 steps to prepare, setup and use a simulation:
· First, the real-world problem is analyzed.
· Second, the modelling objectives are set up through conceptual modelling, which indulges abstracting a model from a real or proposed system.
· Third, the level of detail, scope and goals of the simulation are set up.
· Thereafter, a computer model can be coded.
· Finally, solutions can be derived from the computer model, which can be implemented in the real-world again.
The simulation environment gives us at least two use cases where simulation is beneficial:
a) It enables us to optimize the current, already existing processes
b) It enables us to evaluate the design choices for a conceptual plant or processes.
Manufacturing simulation can help to evaluate with amongst others the following aspects:
· Facility layout and resource allocation
· Throughput and capacity planning
· Management of inventory levels and replenishment rates
· Production planning, scheduling, production line balancing
· Production logistics and material flow
· Process improvement, bottleneck analysis
· Predictive maintenance
Methods in Simulation Modeling
There are at least five main methods in simulation modeling (Figure 2):
· Agent-Based — Agent Based Simulation (ABS) is modelled by analyzing agents that interact with each other. These simulations are typically used when a human (agent) factor is involved. For example, airports, supermarket queues, assembly lines, etc.
· Discrete-Event — Discrete Event Simulation (DES) is a method to simulate real world systems through discrete time steps e.g., events. On every event and process is simulated and is assigned with a specific timestamp. With smaller time steps, meaning more events, the simulation can be more accurate. However, modeling with smaller time steps increase the calculation time of the simulation.
· System dynamics — a System Dynamics Simulation (SDS) method which describes the system and the influence of the environment upon it at the macro level, based on cause-and-effect relationships, reaction delays and feedback loops. The range of SD applications includes also urban, social, ecological types of systems. In SD, the real-world processes are represented in terms of stocks (e.g., of material, knowledge, people, money), flows between these stocks, and information that determines the values of the flows.
· Dynamic systems — a method that is used to describe complex dynamical systems making use of differential equations or partial differential equations . Simulating these systems solve the state-equation to see if the behavior of the system remains stable over time. These systems can typically be found in Control Engineering.
· Hybrid modelling (ABS and DES) — a hybrid approach of combining ABS with DES. This type of simulation can be extremely powerful in situations where processes need to be modelled with stochastics involved. For example, an assembly line production process where raw materials parts need to be assembled produced to a final product. Discrete Event Simulation can model the processes and simulate the accuracy of the assembly line production line for each time step. The agents can be used to model parts, orders and operators, where each agent has their specific set of parameters that are predefined.
Levels of Abstraction and Types of Simulation on Abstraction Level
Figure 3 below shows the set of problems that can be addressed with simulation modeling. These problems are arranged on a scale of level of abstraction and details. In higher abstraction level, problems are analyzed in strategic level and detail level of the model is low. Individual elements such as people, parts and products are not considered at this level. If we look at micro level, then we have the opposite, individual elements are modeled, every single detail can matter. Manufacturing simulation falls under medium detail level because instead of modeling every station or conveyor individually, generally average timings are used in modeling.
The simulation methods mentioned before are also shown in the same figure. Technically, Discrete Event and Agent Based simulations deal with mainly discrete time, while System Dynamics and Dynamic Systems deal with continuous processes. System Dynamic is applied where abstraction level is highest. For factory floor and supply chain processes Discrete Event Simulation is best used. Agent Based method is best used from low to high level of abstraction. These simulation methods are not exclusively used in the Manufacturing industry. Other industries have similar problems that can be tackled with simulation.
Comparing simulation software
Creating a simulation environment can be done in many different coding environments. Büth et al (2017) classify the simulation software according to their range of applications and simulation methods they are capable of. Figure 4 shows a classification of simulation software, starting with general programming languages like Python, C++ or Java that offer the most versatility. But they are also the most time consuming to use compared to simulation programs. General purpose simulations tools offer a wide range of capabilities but may have a longer learning curve. Specific manufacturing simulations focusses on the manufacturing process, serving a better purpose in that area, but are more restricted when external factors are implemented. Manufacturing oriented software can distinguish general manufacturing and one specific application.
· Arena (Rockwell Automation) — is a DES software that enables manufacturing organizations to increase their production throughput, identify process bottlenecks, improve logistics and evaluate potential process changes. It allows to create digital models of the manufacturing, packaging and logistics.
Use case — Was used to evaluate the new manufacturing line for an automobile maker. Another case study was for a major manufacturer of household appliances, it enabled the appliance manufacturer to redesign the assemble process.
· Anylogic — A simulation tool based on Java code that can cope with three types of simulations: DES, ABS and SDS. Furthermore, Anylogic can be used for Hybrid modelling. It has multiple libraries for process modelling, fluids, rail, pedestrian, road traffic and material handling. Anylogic can be used for optimization experiments and Deep Learning with Microsoft Bonsai.
Use case — Model the supply chain of a large steel manufacturer in Europe, where the main objective was to increase the reliability of the plant.
· Tecnomatix Plant Simulation (Siemens Product Lifecycle Management) — uses DES method and enables the modelling and simulation of manufacturing, warehousing and logistics processes. It can be used to explore system’s characteristics and optimize its performance. It allows the analysis of material flow, resource utilization and logistics for all levels of manufacturing planning from global production facilities to local plants and specific lines. It is also a digital twin modeling environment where a virtual model of the plant can be linked to real plant control.
Use Cases — Siemens Energy used this software to assist them in planning for the new production plant. GKN Aerospace and German automotive industry are among other companies that use this tool to optimize production processes.
· Simul8 — a simulation tool that can be used for general purposes and works with DES, ABS and SDS. Furthermore, Hybrid modelling is also included in Simul8. The simulation software is used in almost all sectors, such as, manufacturing, healthcare, defense and many more.
Use case — Increase the production throughput for synthetic Christmas trees by 50 percent and a forecast over a 5-year period was made with Simul8.
· Flexsim — a tool that can be used for the simulation of DES, ABS and SDS. This tool is like Simul8 and is also used in many industries.
Use case — Making a strategic planning tool to make decisions based on capacity issues and supply chain costs for manufacturing sites globally.
· Simio — a tool with the same capabilities as Anylogic, Simul8, Flexsim.
Use case — Making a simulation model to determine the staffing strategy and warehouse capacity for a distribution center.
· Prespective – Provides a digital twin software for Industry that enables to create real-time digital prototypes of machines and systems.
Use case — Implemented a Digital Twin of VDL Nedcar’s production line using the power of Real-time 3D software.
Our experience in applying simulation tools
Anylogic in automotive industry
We designed a simulation in Anylogic for a client to increase the throughput of a process to meet the increasing demand. The goal of this simulation was to gain insight in how the KPI’s are affected by different stochastics. First, a baseline model with fixed parameters was set up. Thereafter, different configurations were made to simulate how the KPI’s were affected. Finally, the different parameters were changed to optimize the KPI’s. The results of these simulations gave the plant a blueprint to increase their throughput.
Tecnomatix in the chemical industry
Another case of simulation project was for a global chemical company in Netherlands. The goal of the project was to evaluate the performances of alternative supply chain configurations according to the determined KPI’s. In the baseline model, current supply chain configuration was modeled in Tecnomatix Plant Simulation software and this model was used for validation. Afterwards, alternative configurations were modeled in the simulation software and the performances of these scenarios were compared versus the baseline model. It helped the plant to analyze the bottleneck in the current situation and the best performing (in relation to KPI’s) alternative was chosen to be implemented.
Conclusion
There are four at least five main methods types of simulation that exist, where DES, ABS and Hybrid modelling are most common in the manufacturing industry. Following the trend of digital manufacturing, virtually test manufacturing processes with simulation makes simulation the digital twin for the manufacturing industry. Simulation software can be used to predict and optimize the performance of different scenarios/solutions before implementing them in real life. For example, one can forecast the utilization of a factory based on historical data. A wide number of applications is available for plant simulation, from generic programming languages as dedicated manufacturing plant simulation. One must choose a simulation application that fits their specific needs, which is also depending on using the simulation for evaluation or optimization.
From our own experience we know that setting-up a well-defined conceptual model is critical for an accurate simulation. We built several models and ran simulations for different plants, which enabled us to optimize the current processes, identify bottlenecks, run what-if scenarios to assess the alternative designs.
We have seen that in most cases simulation models are made for a specific project and upon completion, these simulations are not used anymore. However, the real power of its capabilities lies within the integration to a digital twin. Therefore, our vision is to combine all models into one big model where every separate simulation can be connected to. Implementing plant simulation to a digital environment where the real time data is gathered will unlock even further capabilities, such as real time monitoring, optimizing processes from actual data, better predictive maintenance forecast and so on. Integration of plant simulation into the digital twin falls in line with the ongoing trend of digital transformation journey and thus, is the way forward for smart manufacturing.
References
1. Borshchev and A. Filippov, “From System Dynamics and Discrete Event to Practical Agent Based Modeling: Reasons, Techniques, Tools”, The 22nd International Conference of the System Dynamics Society, July 25–29, 2004, Oxford, England. http://www2.econ.iastate.edu/tesfatsi/systemdyndiscreteeventabmcompared.borshchevfilippov04.pdf
2. L. Büth, N. Broderius, C. Herrmann and S. Thiede, “Introducing agent-based simulation of manufacturing systems to industrial discrete-event simulation tools,” 2017 IEEE 15th International Conference on Industrial Informatics (INDIN), Emden, Germany, 2017, pp. 1141–1146, doi: 10.1109/INDIN.2017.8104934. https://www.horizon2020-perform.eu/files/documents/INDIN%C2%B417_Introducing%20agent-based%20simulation%20of%20manufacturing%20systems%20to%20industrial%20discrete-event%20simulation%20tools.pdf
3. S. Robinson, “Conceptual modelling for simulation part i: Definition and requirements” Journal of the operational research society, vol. 59, no. 3, pp. 278–290, 2008. https://cyberleninka.org/article/n/1101765.pdf
