Development of an intelligent Digital Twin for prediction and control of the process behavior via transient flowsheet simulation using the example of fluidized bed spray granulation

尽管颗粒技术领域的工艺已经在工业中使用了几十年，但为了避免不必要的工艺状态和不合规格的生产费用，生产工厂的过程监控和控制是随着对新发展的持续需求而出现的。该转移项目的主要目的是通过开发智能数字孪生来提高化学工程过程的效率。这将使用流化床喷雾造粒过程的例子进行演示。数字孪生再现了物理工厂的状态，并预测其未来的行为，实现了过程的可靠控制。数字孪生的智能是由底层模型定义的，它将通过持续的离线和在线分析方法进行全面的数据采集来获得。对模型细节的高度关注导致决策建议的实时可用性更快，这允许预测和流程优化步骤。这是作为特定于流程的知识模块开发的，该模块调用流程图模块，校准和整合数据。仿真结果由软传感器解释。将开发一个通信网关，负责仿真结果和数字过程图像之间的交互。流化床喷雾造粒模型将在汉堡工业大学的一个实验室规模的工厂进行实验，应用实验的统计设计。有意调整不需要的工艺条件，以确定稳定工艺运行的边界。结果模型在流程图仿真软件Dyssol中实现，该软件是在DFG前身项目（SPP 1679）期间开发的。通过应用该软件工具，数字孪生体的开发过程可以轻松地转移到固体过程工程的其他过程和过程链中。基于模拟得出的未来过程状态，行动建议将实施到工厂的过程控制系统中。优化算法确定那些导致稳定运行的参数。过程控制系统的实施是在工业合作伙伴Pergande集团公司进行的。除了实验室规模的工厂的操作，转移到更大的工厂将被证明。现有辅助系统的扩展将为新工厂的分布带来竞争优势，并且是工业合作伙伴在流化床技术领域的独特功能。此外，假设由于工厂的可用性提高，合同制造领域的效率可以提高约15%。

The process monitoring and control of production plants comes along with the ongoing demand for new developments to avoid unwanted process states and sub-specification production charges even though the processes in the field of particle technology have been used in industry for many decades. The main aim of this transfer project is increasing the efficiency of processes from chemical engineering by the development of an intelligent Digital Twin. This will be demonstrated using the example of the fluidized bed spray granulation process. The Digital Twin reproduces the state of the physical plant and predicts its future behavior, enable reliable control of the process. The intelligence of the Digital Twin is defined by the underlying models, which will be obtained by comprehensive data acquisition with continuous offline and online analysis methods. The high attention to detail of the models result in decision recommendations available faster the real-time, which allows predictions as well as process optimization steps. This is developed as a process-specific knowledge module that calls a flowsheet module, calibrates and consolidates the data. The results of the simulations are interpreted by soft sensors. A communication gateway will be developed that is responsible for the interaction between the simulation results and the digital process picture. The fluidized bed spray granulation model will be developed by experiments with a lab-scale plant at Hamburg University of Technology by application of statistical design of experiments. Unwanted process conditions will be deliberately adjusted in order to define the borders of stable process operation. The resulting model is implemented in the flowsheet simulation software Dyssol which has been developed during the DFG predecessor project (SPP 1679). By applying this software tool, the development procedure of the Digital Twin can be easily transferred to other processes and process chains of solids process engineering with minimal effort.Based on the future process states derived by the simulations, recommendations for actions will be implemented into the process control system of the plant. An optimization algorithm determines those parameters that result in a stable operation. The implementation into the process control system is conducted at the industrial partner, the company Pergande Group. Besides the operation of lab-scale plants, the transferability to larger plants will be demonstrated. The extension of current assistant systems will result in a competitive advantage for the distribution of new plants and is a unique feature in the field of fluidized bed technology for the industrial partner. In addition, it is assumed that the efficiency in the field of contract manufacturing can be increased by about 15 % due to a higher availability of the plants.