The basic idea of a “Digital Factory Twins” is to virtually mirror real production processes using simulation models. While well-known concepts primarily reflect the resource and parts status, new concepts that cover the individual characteristics of each production part open many more possibilities.
During the production preparation phase, it is state of the art to simulate the process considering the most relevant parameters. Statistical analysis is common to test the robustness of a single production process, but the tolerance ranges across multi-level production processes are largely ignored. This is because combinatorial analysis is far too complex and would probably show that a worst-case combination simply would not work. The solution is to narrow the tolerances of the parameters to meet the functional requirements under all conditions. Unfortunately, this is the decisive driver for the manufacturing costs.
A possible way out of this dilemma could be to ensure compliance with the functional requirements of each component produced already during the production process rather than meeting the (indirect) tolerance and process requirements. This approach requires the analysis of the function (via a simulation), with their relevant and individual properties in parallel to the current process, and preferably within the real cycle time. As a result, every component produced is linked to its production parameters and thus has its own individual fingerprint characteristics. Such an approach is superior to conventional production planning if there is not enough experience or data to provide accurate parameter settings for the equipment, especially in small batches or sample productions, scenarios of lot size 1 or in the ramp up phase of mass production.
Linking process and function simulation
A hot stamping production process as an industrial reference process is shown in Figure 1 and illustrates the concept of a multi-stage production process. The real production process is accompanied by a virtual process. The virtual process is temporally synchronized with the real process so that sensor data of the machines, the process or the environment can be used directly by the process model as parameters in real time. This synchronization enables continuous validation of the virtual process models. For the overall system, it does not matter if the process models are inherently numeric (eg, FEM) or analytic (eg, replacement models). Active control of real processes can be integrated into the virtual process model in an identical way through co-simulation in real time.
Figure 1: Hot stamping process as a reference use-case of a function-oriented process control – Instead of controlling the punch speed over the part temperature (indirect parameter), the process uses the actual yield strength, eliminating the need for additional tolerance chains
The functional evaluation of the produced component is carried out virtually based on the now existing digital component twin (“Digital Twin”) with a functional model. The Digital Twin has all functionally relevant properties, including those that cannot be measured on the real component in a timely, nondestructive or cost-effective manner.
Thanks to the evaluation results from the virtual part assessment vs. the EoL / CoP check is translated with the function-oriented process model control parameters and transferred to the machine controls via the function-oriented process control. This ensures that each individual part produced meets the functional requirements, even if the replacement test at the EoL / CoP station cannot fully attest to this, e.g. if destructive testing would be required.
CI/CD approach to foster Digital Factory Twin approach
A key requirement for implementing the concept in a smart factory environment is a lean management of the required simulation models. Process models provided by production planning departments and functional models brought in by engineering are managed in cloud repositories that also implement lifecycle management. To ensure interoperability within (real-time) co-simulation, the models involved must comply with Distributed-CoSimulation-Protocols (DCP) standards. Real-time data streams from IoT sensors or machine data are connected to the co-simulation service via DCP slaves.
Figure 2: CI/CD concept to enable a lightweight and low investment factory environment – based on state-of-the-art communication standards (DCP, OPC-UA) and model repositories, production processes are mirrored to ensure conformity
Ideally, after the parameters have been settled and the models fitted and parameterized, a local and lightweight numerical model to control the production process can replace the cloud solution.
The application possibilities for the individualization of production components with their specific properties concept are manifold. In addition to trend analysis of component properties, predictive plant control during operation, and predictive maintenance, measures can be taken to verify in real time the effectiveness of the monitored process parameters or component function. In addition, with valid virtual components, it is possible to perform virtual commissioning at the plant manufacturer in advance without the poorly available real components.