No matter how well designed your production plants are – if your production schedules are not top-notch, you will lose a lot of time and money. Using digital twins by INOSIM, the OptiProd.NRW project is developing a new methodology and software tools that let you generate high-quality production schedules with the click of a button even for the most complex of plants. And these schedules will just work – no need for expensive, error-prone, and time-consuming manual modifications.
A New Dimension In Production Scheduling
In the first project year, the consortium has made large strides towards this ambitious goal. Partners TU Dortmund and INOSIM have built a full-scale prototype of the new software system for optimal production scheduling. The system is based on a new optimization approach that combines a tailored evolutionary algorithm with customized heuristics that are able to identify the most promising candidate solutions and that intelligently reduce the search space so that high-quality schedules can be found even for the most complex plants in reasonable time. And this algorithm now also supports campaign scheduling – it is able to automatically determine the best way to split your product demand into production batches so that your objectives and constraints are always met.
Massively parallel simulation engine
The software system includes a massively parallel simulation engine for INOSIM models – a key requirement for the efficient computation of optimal production schedules. The engine distributes simulation requests automatically to a large number of INOSIM instances and returns the results to the optimizer. The engine is currently running on high-performance servers at TU Dortmund and at INOSIM. Currently, we are achieving a 20-fold performance increase of the optimizer compared to single-instance simulation. And we are working on pushing the performance much further in the future.
Demonstrating the power of the new methodology and software system on truly industrial-scale use cases is a key goal of the project. We are about to finish building a detailed INOSIM simulation model of a complex industrial formulation plant by partner Bayer that also includes the associated logistics processes (implemented in the new packaged-goods library of INOSIM). The model represents the plant so faithfully that we expect the generated schedules to directly be applicable at the real plant. In addition, our master students Robin Ehrhardt and Engelbert Pasieka have built simulation models of other complex, more academic use cases that allow us to test the new tool on a variety of different scheduling challenges that regularly occur in industrial practice.
Full-Scale Industrial Use Case
Fueled by the swift progress in the first project year, all project partners are looking forward to continuing their work over the next months and years. We expect that we will be able to demonstrate the new system on the full-scale industrial use case in the near future, and we are hard at work to improve and generalize the prototypical implementations of the methodology and software tool into a mature solution for optimal production scheduling during the remainder of the project.
For more details on the OptiProd.NRW project, please visit the project website.