Using digital twin technology in production planning and control process: An application in textile industry

Today, with the combination of existing production and design technologies with modern information and communication technologies, the digitalization process in production and design has begun. Especially with the emergence of Industry 4.0 technologies such as the internet of things, cloud computing, big data analytics, artificial intelligence and digital twin, today's production paradigm has begun to transform into intelligent production. One of these technologies, the digital twin, creates a virtual and real-time representation of production systems or components, providing future-oriented information about products or production systems with constantly collected data. In addition, it provides many advantages such as detecting problems in processes, speeding up the design cycle of products, identifying the source of quality problems and preventing errors. Because of these advantages, it has started to be applied more and more in areas such as smart manufacturing, smart building management, smart city, health, oil, gas and many more. However, in order to create and implement digital twins of products or systems, first of all, necessary data must be collected, organized, processed and made meaningful. In this study, it is aimed to examine the processes from end to end within the scope of the data digitization project in the production facility of a leading company operating in the garment industry and to make data-oriented process designs using new generation information technologies. Accordingly, it is targeted to create the necessary process infrastructures for making digital twin models, which is a newly developing and rapidly growing technology. In the study, first of all, process maps were created and the constantly changing data of the processes were obtained with the help of sensors and interfaces and transferred to the system. Then, by establishing a connection between the process-based times taken from the machines on the production line and the characteristics of the product to be produced, how long it will take for any product to be completed when it enters the process was estimated on the Knime platform using linear regression, polynomial regression, gradient boosting decision forest regression and random forest regression algorithms. According to the estimation results, it was determined that the random forest regression model had the lowest error metric values and this regression model was integrated into the ERP infrastructure. With estimation, standard times were obtained for products on the basis of classification and supply, and costing, productivity measurement, reporting and performance evaluation studies were carried out with these standard times. In addition, the production scheduling study has been designed according to the estimated production times and various parameters on the line and will be tested by the company. The study is important in terms of creating the infrastructure of a smart system that can decide on its own, and it is thought that it will contribute to the creation of the digital twins of the processes.