Estimation of Operation Time with Digital Twin in Manufacturing

Abstract

There are two most important factors that are taken into consideration in businesses. One of them is time and the other is cost. In order to save time and cost, planning of production subcomponents involves a series of critical activities. Making more effective plans in the field of production has become possible with industry 4.0. Industry 4.0 includes the digitalization of production. One of the most popular topics in this process is the Digital Twin [DT]. The DT philosophy has enabled businesses to better understand the sub-processes of production. In this way, they can optimize them. With the development of this philosophy, more detailed models have been created. Enterprises keep their data under control in order to control, manage and optimize processes. These data are then utilized in the model building process. The aim of this study is to estimate the time from the moment a product enters the process to the moment it leaves the process by using the data obtained through time study etc. studies. Automated machine learning (AutoML) method is used to build the best model. Machine learning (ML) algorithms, which are popularly used in the literature, may not always give the best result. In order to prevent this, starting from the data preprocessing step, including hyperparameter optimization, the aim is to find the algorithm and parameters that give the best performance. It will contribute to DT studies by estimating the operation time. The study used a 115-row dataset from CNC machines. The dataset consists of velocity, motion and actual time. The actual time is tried to be estimated using motion/speed. It is aimed to achieve the best results with AutoML. lazy predict and tpot library were used in the study. As a result, an estimation of the duration of 100% was realized.